Coder Team 

Coder, the leader in self-hosted AI development infrastructure for the enterprise, today announced the beta release of <a href="http://coder.com/blog/introducing-coder-agents">Coder Agents</a>. The solution is a native AI coding agent designed to run entirely on self-hosted infrastructure and empower developers to use any AI model they desire, giving enterprises the security and governance required to run AI-driven development workflows at scale.

<img src="https://www.datocms-assets.com/19109/1778013320-pr_coder-agents.png" alt="Coder Sets a New Standard for AI Coding with Self-Hosted, AI Model Agnostic Coder Agents">

AI adoption in software development has accelerated rapidly, with <a href="https://coder.com/blog/what-100-engineering-teams-revealed-about-ai-maturity-and-what-to-do-about-it">61% of engineering teams already running agents</a>. Yet most remain in early stages of maturity and lack the infrastructure to scale them safely. Many rely on tools with cloud-hosted orchestration, sending code and context to third-party services with limited visibility or control. For large enterprises and companies in highly regulated industries, this tradeoff is unacceptable. 

Built directly into Coder, Coder Agents gives enterprises a way to deliver modern, agent-driven developer workflows without sending source code, prompts, or model interactions outside their network perimeter. Instead of running agents in a vendor-controlled cloud, the entire agent system — including the control plane, orchestration, and execution — runs on infrastructure owned and operated by the customer, enabling centralized governance and secure, scalable development across the organization.

“Companies are being forced to choose between adopting AI agents and maintaining control over their infrastructure and data,” said Rob Whiteley, CEO at Coder. “Coder Agents removes that tradeoff. You get a modern, conversational agent experience, entirely inside your own environment. Now enterprises can equip any employee with any AI coding model of their choosing, with the governance and flexibility they require.”

<h2>The standard for AI coding workflows</h2>

AI coding workflows are entering a new phase. As model capabilities continue to advance, differentiation is shifting away from the agent itself and toward the infrastructure required to run it securely and at scale. Despite this demand, <a href="https://coder.com/blog/what-100-engineering-teams-revealed-about-ai-maturity-and-what-to-do-about-it">Coder research</a> shows that 70% of companies are deploying agents on infrastructure that was never designed to support them, exposing a gap between adoption and enterprise readiness.

Coder addresses this challenge, enabling organizations to:
* Keep source code, prompts, and model traffic within the network boundary
* Run agents in fully air-gapped or network-restricted environments
* Connect to any AI model provider or self-host AI models entirely without intermediary routing
* Enforce centralized policies for model access, prompts, and usage across teams

For platform teams, Coder Agents provides a standardized way to deploy and manage AI agents across the organization. Instead of fragmented tooling and inconsistent configurations, teams can centralize model access, enforce policies, and gain visibility into how agents are used and what they produce. Developers can delegate tasks such as writing code, generating tests, analyzing repositories, and opening pull requests through a conversational interface and API.

Coder Agents is available now in beta, with full feature access and no usage-based limits through September. <a href="https://coder.com/docs/ai-coder/agents/getting-started">Learn more.</a>

<h2>Coder Agents Frequently Asked Questions</h2>

<h3>Q: What is Coder Agents?</h3>

A: Coder Agents is a native AI coding agent that runs on self-hosted infrastructure. The agent operates from the Coder control plane and provisions workspaces only when code execution is required, allowing developers to complete tasks like writing code, generating tests, and opening pull requests without sending source code or prompts outside their environment.

<h3>Q: How is Coder Agents different?</h3>

A: Most AI coding agents rely on cloud-hosted orchestration, where parts of the agent workflow (planning, inference, or coordination) run on vendor infrastructure. Coder Agents runs the entire agent system, including the control plane, orchestration, and execution, on infrastructure owned by the customer. This allows organizations to maintain full control over data flow, enforce governance policies, support air-gapped environments, and choose any model provider without intermediary routing.

<h3>Q: How is Coder Agents different from tools like Claude Code, Codex, or Cursor Agents?</h3>

A: Coder Agents is not a wrapper around tools like Claude Code or Codex. It uses its own native agent and runs entirely on customer-controlled infrastructure. Unlike many AI coding tools that rely on cloud-hosted orchestration, Coder Agents keeps the full agent system (planning, orchestration, and execution) inside the customer’s environment. This provides greater control over data, governance, and model usage.

<h3>Q: Why do enterprises need self-hosted AI coding agents?</h3>

A: Enterprises, especially in regulated industries, need to control where source code and model interactions are processed. Many existing AI coding tools require sending code and context to third-party cloud services, which creates challenges around data residency, compliance, and auditability. Self-hosted agents allow organizations to adopt AI-driven development while keeping all activity within their own infrastructure, ensuring security, governance, and visibility.

<h3>Q: Does Coder Agents lock teams into a specific AI model?</h3>

A: No. Coder Agents is model-agnostic and supports multiple providers, including Anthropic, OpenAI, Google, AWS Bedrock, and self-hosted models. Platform teams can centrally control which models are available, and developers can use approved options without being locked into a single vendor.

<h3>Q: Can Coder Agents run in air-gapped or regulated environments?</h3>

A: Yes. Coder Agents is designed to run entirely on self-hosted infrastructure, including cloud VPCs, on-prem environments, and fully air-gapped deployments. Organizations can also connect to self-hosted models, ensuring that no code, prompts, or model traffic leaves their network. This makes it suitable for industries with strict security, compliance, or data residency requirements.

New native agent architecture enables enterprises to run AI-driven development workflows with full governance over infrastructure, data, and models

Coder Sets a New Standard for AI Coding with Self-Hosted, AI Model Agnostic Coder Agents

Coder, the leader in self-hosted AI development infrastructure for the enterprise, today announced that it has been ranked 15th on the <a href="https://www.inc.com/regionals/southwest">Inc. 5000 Southwest regional list</a> of fastest growing private companies. 

An extension of the national Inc. 5000 list, the Regionals list offers a data-driven look at the independent small businesses driving growth across the Southwest economy. Coder was listed for its strong revenue and employee growth over the past two years. 

Coder provides the infrastructure enterprises need to securely run and scale AI-powered software development. Its self-hosted platform enables developers and AI agents to work in isolated, policy-governed environments with built-in security and control. This allows organizations to operationalize AI safely while improving speed, efficiency, and cost. 

“Ranking among the fastest-growing companies in the Southwest is a strong signal of where the software industry is headed,” said Rob Whiteley, CEO of Coder. “As organizations scale AI-driven software development to all employees, they’re realizing that traditional infrastructure can’t support the demands of autonomous agents and modern AI tools. Our growth reflects a broader shift toward secure, enterprise-controlled environments where anyone building software can operate AI with speed and accountability.”

Coder, along with the other companies on this year’s list, demonstrate exceptional revenue expansion, resilience, and job creation during a challenging economic period. Between 2022 and 2024, these 132 private companies had a median growth rate of 79 percent; by 2024, they’d also added 9,633 jobs and $5.2 billion to the region’s economy. 

This ranking comes on the heels of Coder&#39;s <a href="https://coder.com/blog/90m-series-c-led-by-kkr-to-advance-secure-enterprise-ai-development">$90 million Series C funding round that was announced last week</a>.

An extension of the national Inc. 5000 list, the Regionals list offers a data-driven look at the independent small businesses driving growth across the Southwest economy. Coder was listed for its strong revenue and employee growth over the past two years.

Inc Regionals Southwest

Coder Ranks 15th on the 2026 Inc. 5000 List of the Fastest-Growing Private Companies in the Southwest

Coder Raises $90M Series C Led by KKR: Rob Whiteley on What's Next for AI Development Infrastructure

Coder, a leader in AI development infrastructure, today announced a $90 million Series C led by funds managed by <a href="https://www.kkr.com/">KKR</a>, a leading global investment firm, with participation from Qube Research &amp; Technologies (<a href="https://www.qube-rt.com/">QRT</a>), Uncork Capital, and other existing investors. The investment underscores a shared conviction that cloud-based developer infrastructure will play a critical role in enabling enterprises to securely build and deploy software at scale.

Founded in 2017 in Austin, Texas, Coder provides a centralized, secure <a href="https://coder.com/products/workspaces">platform</a> where enterprises can build and run software development environments in the cloud, replacing fragmented, local setups with standardized, customer-controlled workspaces. By enabling both developers and AI coding tools to operate within the same governed infrastructure, Coder helps organizations improve consistency, accelerate onboarding, and maintain control over how software is built. 

“We chose to partner with customers KKR and QRT in this round because they have the clearest view of how AI evolves software development,” said Rob Whiteley, CEO of Coder. “They’re using Coder to bring consistency and security for every enterprise user to leverage the latest technologies like Claude Code, Cursor, and OpenClaw. Together, we are rearchitecting the foundation for how enterprise software is built.”

“Software development is undergoing a fundamental shift, driven by the move to the cloud and the rapid adoption of AI, with more than 80% of enterprise developers today using or planning to use coding agents in their daily workflows,” said Ben Pederson, Managing Director at KKR. “As enterprises scale the use of AI in development, they need infrastructure that allows that work to happen in a secure, standardized, and repeatable way. We believe Coder is well positioned to serve as the foundational platform for this next generation of software development.”

The company will use the funding to support platform innovation — with a focus on supporting enterprise AI workflows and strengthening <a href="https://coder.com/solutions/ai-governance">governance</a> capabilities — and to scale its presence in Europe, Asia, and North America to meet demand for Coder.

<h2>The new model for enterprise development</h2>

Customers define what successful AI adoption looks like in practice: balancing speed with control, and experimentation with the ability to scale. This moment represents an opportunity to establish a new standard for how to safely and easily access corporate-approved coding tools. Coder automates agentic provisioning, eliminating the need for users to install agents, connect LLMs, provision tools and MCP servers, and plumb libraries, source code repositories, and infrastructure. All of that fades to infrastructure- and policy-as-code that users don’t have to think about.

At KKR, Coder was rolled out to more than 500 engineers as part of a broader effort to align development environments and introduce AI in a governed way. Since first deploying Coder a year ago, the firm has moved from no AI-assisted code to more than half of commits taking place within Coder-managed environments. Based on this initial success, KKR plans to roll out coding tools to many additional functions. The investment will be made primarily through KKR’s Next Generation Technology Growth Fund III.

“Coder has fundamentally changed how we approach software development at KKR, particularly as we integrate AI into our workflows,” said Ruchir Swarup, Chief Information Officer at KKR. “By standardizing development environments and embedding governance directly into the process, we’ve been able to scale the use of AI tools in a secure and controlled way. This has improved consistency across teams, accelerated onboarding, and increased developer productivity and software velocity. We’re now extending these capabilities beyond engineering to analysts, data scientists, and other users.”

Customer-investor QRT experienced similar success. QRT has rolled Coder out to half of its over 2,000 employees, including a mix of software engineers, data scientists, and analysts. Coder is a critical component of how these employees use agentic tooling.

“Deploying agentic AI at our scale requires infrastructure that doesn’t compromise on governance,” said Zohar Melamed, Head of Developer Experience at QRT. “Coder gives us the ability to audit every LLM request, control access across our global infrastructure, and maintain strict compliance while dramatically compressing deployment timelines.”

<h2>Growth led by customer expansion</h2>

Coder initially gained traction by helping enterprises bring structure, safety and control to how software is built. As the scope of software development expands to include AI tools and agents, that foundation is used to support a broader set of workflows across the organization. 

Coder’s growth is accelerating alongside this shift. The company has delivered 300% year-over-year bookings growth over the past four quarters, 45% quarter-over-quarter growth in Q1, and 148% year-over-year growth in Q1. This momentum is driven by expansion within the existing customer base, with net dollar retention at 184%. As enterprises standardize development and adopt AI workflows, Coder consistently expands across teams and use cases, reinforcing its role as a core part of the development stack.

Latham &amp; Watkins LLP served as legal advisor to KKR &amp; Co. Inc.

Visit the company <a href="https://coder.com/blog/coders-customers-led-our-series-c-what-that-tells-me-about-ai">blog</a> to learn more about Coder’s vision for the foundation of safe agentic development.

About Coder

Coder is the leading platform for AI development Infrastructure, enabling enterprises to securely run human and AI-driven development workflows in consistent, governed environments. Coder provides self-hosted, agent-ready workspaces that unify developer productivity and platform governance. With Coder, enterprises can confidently evolve from human-only development to AI-assisted and autonomous workflows—without sacrificing security, compliance, or performance. Learn more at <a href="https://coder.com/">coder.com</a>. 

About KKR

KKR is a leading global investment firm that offers alternative asset management as well as capital markets and insurance solutions. KKR aims to generate attractive investment returns by following a patient and disciplined investment approach, employing world-class people, and supporting growth in its portfolio companies and communities. KKR sponsors investment funds that invest in private equity, credit and real assets and has strategic partners that manage hedge funds. KKR’s insurance subsidiaries offer retirement, life and reinsurance products under the management of Global Atlantic Financial Group. References to KKR’s investments may include the activities of its sponsored funds and insurance subsidiaries. For additional information about KKR &amp; Co. Inc. (NYSE: KKR), please visit KKR’s website at <a href="about:blank">www.kkr.com</a>. For additional information about Global Atlantic Financial Group, please visit Global Atlantic Financial Group’s website at <a href="about:blank">www.globalatlantic.com</a>.

Media contacts

KKR: 
Brooke Rustad or Jenn Bernstein 
<a href="mailto:media@kkr.com">media@kkr.com</a>

Coder: 
Jennifer Tanner 
Look Left Marketing 
<a href="mailto:coder@lookleftmarketing.com">coder@lookleftmarketing.com</a>

KKR is joined by QRT, Uncork Capital, and other existing investors, signaling growing demand for infrastructure that enables the next generation of AI-powered software development 

Coder Series $90M Series C Led by KKR

Coder Secures $90M Series C Led by KKR to Advance Secure Enterprise AI Development

Coder has been named a <a href="https://channelvisionmag.com/channelvision-announces-2026-cvai-award-winners-2/">2026 ChannelVision AI Award winner</a> in the Next-Gen Partner Relationship Management (PRM) category. The award recognizes Coder’s innovative approach to building an AI-enabled partner ecosystem that accelerates the adoption of modern developer infrastructure and supports the growing demand for agentic development.

The CVAI Awards honor technologies that leverage artificial intelligence to deliver meaningful innovation across the channel. Coder was recognized for its AI-enabled partner engagement model, which helps integrators, consultants, and service providers build and scale solutions on the Coder platform while reducing friction across the partner lifecycle.

Through an automated PRM platform, partners gain real-time visibility into opportunities, direct collaboration with Coder teams, and streamlined access to program updates, product announcements, and sales engagement resources. The system integrates directly with Coder’s CRM to surface live deal status, eliminating manual follow-ups and enabling partners to stay aligned throughout the sales cycle.

In addition, partners gain access to Blink, Coder’s AI agent framework, which provides real-time technical and sales enablement. Blink enables partners to quickly access architecture guidance, product knowledge, and deployment insights, which support faster solution development and more effective customer engagement.

“Coder’s AI-assisted partner model reflects the growing shift toward agentic development, where AI agents support developers and platform teams throughout the software lifecycle,” said <a href="https://www.linkedin.com/in/tim-cleary/">Tim Cleary</a>, head of partnerships at Coder. “This award highlights the effectiveness of our innovative, AI-first approach. By combining secure development environments with AI-enabled partner collaboration, Coder helps organizations adopt modern developer platforms while enabling partners to deliver repeatable, high-value solutions.”

Since launching the program, Coder has processed 50+ new partner opportunities and automated more than 1,800 partner communications, while achieving 65%+ email open rates and 95% engagement on partner announcement threads. This demonstrates the impact of automation and real-time collaboration on partner engagement.

The ChannelVision CVAI Awards recognize organizations that are pushing the boundaries of AI-driven innovation across the channel ecosystem. Visit the website to learn more about how Coder has incorporated AI into its partner strategies. 

<a href="https://coder.com/partners">Learn more about Coder’s Partner Program</a>

Award recognizes Coder’s use of AI and automation to streamline partner collaboration, accelerate co-selling, and enable partners to deliver secure development environments at scale

2026 ChannelVision AI Award

Coder Wins 2026 ChannelVision AI Award for Partner Relationship Innovation

Coder announced today that it has achieved the Amazon Web Services (AWS) DevOps and the Generative AI Competencies in the Agentic Tools Category. This specialization recognizes Coder as an <a href="https://coder.com/partners/aws">AWS Partner</a> that helps customers and AWS Partners drive the advancement of services, DevOps tools, and infrastructure pivotal for implementing generative AI technologies.

Achieving these Competencies differentiates Coder as an AWS Partner that has demonstrated technical proficiency and proven customer success, powering secure, consistent environments where humans and AI code together, trusted by leading enterprises worldwide. Coder possesses the experience and DevOps expertise shown from successful projects for addressing customer challenges using generative AI solutions as an enabler of their digital transformation strategies for augmenting the customer experience, delivering hyper-personalized and engaging content, streamlining workflows, and delivering actionable results powered by generative AI technology from AWS.

&quot;Enterprises need infrastructure that enables secure AI development at scale without compromising governance or control,&quot; said Rob Whiteley, CEO of Coder. &quot;Achieving the AWS Generative AI and DevOps Competencies validates Coder’s commitment to helping enterprises and government agencies adopt AI-driven development workflows with the security, observability, and compliance controls they require. We&#39;re focused on making it easier for organizations to build on AWS with confidence.&quot;

The <a href="https://aws.amazon.com/partners/programs/competencies/">AWS Competency Program</a> aims to assist customers in connecting with AWS Partners who possess extensive knowledge and technical expertise in using AWS technologies and best practices. These Partners facilitate the seamless integration and deployment of solutions to meet the unique needs of both startups and global enterprises.

Achieving the AWS Generative AI and DevOps Competencies reflects Coder’s focus on enabling secure, observable, and practical AI adoption in enterprise software development. <a href="https://coder.com/docs/ai-coder/tasks">Coder Tasks</a> allows teams to run long-lived AI agents in dedicated, isolated workspaces, reducing repetitive development work such as bug fixes, documentation, and QA while giving agents access to real source code, developer tools, and GitHub context under enterprise controls. <a href="https://coder.com/blog/launch-dec-2025-agent-boundaries">Agent Boundaries</a> adds network isolation and detailed logging to limit risk from unpredictable agent behavior, preventing data exfiltration, prompt injection, and unsafe command execution without relying on brittle sandboxes or constant human approval. <a href="https://coder.com/blog/launch-dec-2025-ai-bridge">AI Bridge</a> centralizes all AI usage across tools, agents, and models, providing identity-aware access, audit logs, and cost visibility while reducing API key sprawl and enabling organizations to govern and measure AI use consistently across teams.

This specialization recognizes Coder as an AWS Partner that helps customers and AWS Partners drive the advancement of services, DevOps tools, and infrastructure pivotal for implementing generative AI technologies.

Coder + AWS

Coder Achieves AWS Generative AI and DevOps Competencies

Coder, the leading platform for AI Development Infrastructure, enabling missions to securely run human and AI-driven development workflows in consistent, governed environments, today announced the listing of its Coder Premium AI Developer Infrastructure solution in the AWS Marketplace for the U.S. Intelligence Community (ICMP). ICMP is a curated digital catalog from <a href="https://coder.com/partners/aws">Amazon Web Services (AWS)</a> that makes it easy to discover, purchase, and deploy software packages and applications from vendors that specialize in supporting government customers.

Federal agencies are under increasing pressure to accelerate software build cycles, and adopt AI-assisted development workflows while maintaining strict security, compliance, and data sovereignty requirements. To meet these demands, agencies need a modern, secure, self-hosted development infrastructure platform like Coder that equips mission builders with centrally governed, high-performance development workspaces—enabling rapid capability delivery with embedded policy controls, guardrails, and comprehensive audit trails to ensure mission-critical workflows execute with full assurance across all classification levels.

With this foundation in place, agencies can accelerate mission software development, where 10x build velocity becomes a force multiplier, with comprehensive governance, modern workflow experiences, and end-to-end audit of all human and AI builds processes as a standard of their operating basis. They can also adapt rapidly by introducing new tools and version updates in minutes with less operating overheard using immutable terraform template controls across thousands of users, while optimizing infrastructure utilization to reduce waste, rationalize tooling, streamline container image “sprawls”, and return compute and human time to priority mission outcomes.

“Government engineering teams need a way to adopt AI-assisted development without sacrificing control over their infrastructure, data, or security posture,” said Rob Whiteley, CEO of Coder. “By listing Coder in ICMP, we’re making it easier for federal customers to adopt a modern development environment platform that aligns with their security, compliance, and operational requirements.”

Coder is used today across all impact levels by those who serve and the <a href="https://coder.com/partners">system integrators</a> that support them globally to deliver meaningful outcomes across a wide range of programs and missions. 

Visit the website to learn more about how Coder works to support federal customers at <a href="https://coder.com/solutions/government">coder.com/solutions/government</a>.

Coder, the leading platform for AI Development Infrastructure, enabling missions to securely run human and AI-driven development workflows in consistent, governed environments, today announced the listing of its Coder Premium AI Developer Infrastructure solution in the AWS Marketplace for the U.S. Intelligence Community (ICMP). 

Coder Listed in AWS “ICMP” for the US Federal Government 

April 2026 update: AI Bridge is now AI Gateway, and Agent Boundaries is now Agent Firewall, with no functional changes. This post also references Coder Tasks, an earlier approach to orchestrating agent workflows on Coder infrastructure. For current capabilities, see <a href="https://coder.com/docs/ai-coder/agents">Coder Agents</a>.

<hr>

Today, Coder introduced new capabilities that bring AI coding agents into secure, self-hosted workspaces, enabling enterprises to adopt AI safely at scale. <a href="https://coder.com/blog/launch-dec-2025-ai-bridge">AI Bridge</a>, <a href="https://coder.com/blog/launch-dec-2025-agent-boundaries">Agent Boundaries</a>, and major enhancements to <a href="https://coder.com/docs/ai-coder/tasks">Coder Tasks</a> deliver industry-first capabilities that bring security, observability, and control to hybrid development environments where humans and AI agents work together.

Enterprises are racing to adopt AI, but most lack the necessary infrastructure for success. A recent Cisco study found that only 13% of global companies have a defined AI strategy. The rest rely on fragile workarounds such as running coding agents locally, ad-hoc isolated sandboxes, unmanaged key sharing, compliance gaps, and GPU-heavy workflows that do not scale. Developers, data scientists, and other builders face similar friction: broken environments, identity sprawl, and inconsistent access to data and tools. What’s missing is a secure, governed foundation that treats both developers and agents as first-class users, providing consistent context, control, and efficient compute across every workflow.

Coder solves this problem by extending the company’s <a href="https://coder.com/cde">open source, self-hosted development platform</a> into a <a href="https://coder.com/cde/premium">full-stack foundation for governed AI development</a>. With unified policy, identity, and execution controls, Coder enables enterprises to safely evolve from human-only workflows to AI-assisted and autonomous development.

“AI has broken the software development lifecycle. Bolting AI tools onto the old model, where code lives on local laptops, creates risk, cost, and chaos. This gets worse when you add AI agents, which are simply impossible to run concurrently on laptops,” said Rob Whiteley, CEO of Coder. “Coder is transforming the SDLC, making AI development safe, scalable, and production-ready. Now, enterprises have a governed foundation where humans and AI agents can build together with consistent security, identity, and observability.”

<h2>A Unified Governance Stack for Secure, Scalable AI Development</h2>

Coder’s latest release introduces new core capabilities that give enterprises the governance, control, and execution layer required to run AI agents in real development environments.

AI Bridge

AI Bridge: Centralizes access, authentication, and observability for all model providers. It replaces disjointed brittle homegrown proxies to consolidate prompt logs, usage patterns, and token consumption into a single governance plane. Platform teams gain real-time visibility into AI usage.

Agent Boundaries

Agent Boundaries: Provides policy-driven control over what agents can access. This agent-specific firewall enforces explicit allow lists for network destinations, tools, and internal systems, which ensures agents only operate within approved contexts and prevents unintended or unsafe actions.

Coder Tasks: The execution engine that powers both agent-driven and developer-driven automation. Tasks run long-lived, low-interaction jobs with full API + notification support, which is ideal for code review, documentation, issue solving, experimenting, and test authoring. 

Together, these capabilities give enterprises a governed, end-to-end foundation for running AI agents and human developers side by side in secure, self-hosted environments.

Demo: Coder's Features for Efficient and Secure Agent Management

Join Coder’s Launch Week for live demos, technical deep dives, and a hands-on workshop that show how to use Coder’s new AI governance stack in real-world development workflows. Learn more and register at <a href="https://launch.coder.com">launch.coder.com</a>.

Today, Coder introduced new capabilities that bring AI coding agents into secure, self-hosted workspaces, enabling enterprises to adopt AI safely at scale. AI Bridge, Agent Boundaries, and major enhancements to Coder Tasks deliver industry-first capabilities that bring security, observability, and control to hybrid development environments where humans and AI agents work together.

AI Bridge conceptual graphic

Coder Introduces AI Development Infrastructure for Hybrid Human and Agent Teams

Today, Coder introduced Blink in Early Access, which will be unveiled during Cloudflare Connect. Blink is the fastest way for developers to build, ship, and run AI agents and chatbots — starting with Slack.

The demand for AI agents is surging, given their benefits in personalization and efficiency for both businesses and consumers; however, research shows that only a small fraction of in-house agent projects ever reach production status. Skillset gaps and technical obstacles, such as data integration and system complexity, can stall development or lead many organizations to abandon their agent plans altogether. <a href="https://www.reuters.com/business/over-40-agentic-ai-projects-will-be-scrapped-by-2027-gartner-says-2025-06-25/">Gartner</a> warns that over 40% of agentic AI projects will be scrapped by 2027 due to high cost or execution challenges. 

Coder solves this with Blink, an open source CLI for building agents, and Blink Cloud, a hosted service for deploying and sharing them across teams. Unlike existing coding assistants, it’s a full developer engine for building AI agents. It provides the complete runtime, infrastructure, and developer tools needed to build intelligent, event-driven agents that integrate directly with real-world systems. In Early Access, it supports Slack to meet developers where they are. Blink will soon support a deeper integration with GitHub and other environments.

“It’s surprisingly difficult to build and personalize even the most basic AI agents, such as a SlackBot or a GitHub code reviewer,” said Kyle Carberry, co-founder at Coder. “Blink brings AI to the tools where chat already happens and makes development simple: no complex workflow builders, just agents as code.”

<h2>Built for builders</h2>

Open source by design, Blink is built for developers who want practical agents without platform lock-in or overhead. Blink’s open-source agent development framework runs entirely from the terminal, giving developers a fast, code-first way to build and refine agents. Blink enables rapid local feedback loops that mirror the way developers already work. Users can test, adjust, and validate an agent’s behavior in real time before optionally deploying it to the cloud.

Blink Cloud: Deploy and share agents across your team

Blink was created in response to community feedback and insights gleaned from building and supporting Coder’s innovative open source and proprietary AI development infrastructure solutions. This includes the company’s flagship enterprise development environments, which already support AI and tooling workloads with GPU access, modular templates, and secure boundaries for coding agents and developers alike. 

<h2>Launched on the Cloudflare Connect stage</h2>

Today, during Cloudflare’s keynote at Cloudflare Connect, Carberry will join Cloudflare CTO Dane Knecht and Principal Systems Engineer Sunil Pai to highlight how Blink can leverage Cloudflare’s global edge for performant, secure agent deployments. 

What’s available now:

<ul>
<li>Blink CLI: Build and test agents from your terminal</li>
<li>Blink Cloud: Deploy and share agents across your team</li>
<li>Slack support: Create bots that act and respond in Slack</li>
</ul>

Early Access is open and free to download, and feedback and contributions are encouraged. The company invites developers to join them in shaping the future of high-performance agentic software together. 

Start building now at <a href="https://blink.so">blink.so</a>.

Creating a Slackbot in 60 Seconds | Blink Agent Development Engine

Open source solution enables developers to create, ship, and run AI agents from the terminal, without lock-in or overhead

Coder Launches Blink: The Fastest Way to Build and Deploy AI Agents

This post is part of Coder Launch Week (July 7–11, 2025). Each day, we’re sharing innovations that make secure, scalable cloud development easier. Follow along <a href="https://launch.coder.com">here</a>.

<h2>Coder Registry: Improved and open for contributions</h2>

A growing marketplace for secure, enterprise-ready modules—built to share, remix, and accelerate innovation across teams.

<a href="https://registry.coder.com/"><img src="https://www.datocms-assets.com/19109/1752233130-registry-1.png"/></a>

<h3>Challenge: Customization without reinventing the wheel</h3>

Every team’s cloud development workflow is unique, but building that experience from scratch can be slow and error-prone. Whether you&#39;re spinning up workspaces on Kubernetes, integrating with internal tools, or enabling AI agents, it’s not just about provisioning infrastructure. It’s about composing a complete, reliable dev environment.

To make that possible, Coder was built to be composable by design.

At the foundation of this model are:

<ul>
<li>Templates: Terraform-based blueprints that define the infrastructure for a Coder workspace (e.g., EC2, OpenStack, or Kubernetes).</li>
<li>Modules: Add-on components that extend templates with extra functionality (e.g., installing JupyterLab, configuring AI agents, or cloning a Git repo).</li>
</ul>

<a href="https://registry.coder.com/modules"><img src="https://www.datocms-assets.com/19109/1752233130-registry-2.png"/></a>

With just a few lines of Terraform, platform teams can mix and match these components to create powerful, production-ready environments. But until recently, discovering high-quality modules—or publishing your own—was harder than it should be.

<h3>Solution: Major updates to the Coder Registry</h3>

The Coder Registry is the hub for discovering, publishing, and consuming templates and modules. Here’s what’s new:

<ul>
<li>Contributor Guides: Clear documentation and walkthroughs to help you get started fast.</li>
<li>New Governance Process: Submit your module for validation to appear in the public Registry, or publish privately with zero friction.</li>
<li>Stronger Architecture: More resilient backend to support scaling and faster updates.</li>
<li>New Modules including <a href="https://registry.coder.com/modules/coder/jetbrains-fleet">JetBrains Fleet</a>, <a href="https://registry.coder.com/modules/coder/zed">Zed</a>, <a href="https://registry.coder.com/modules/coder/jetbrains">JetBrains Toolbox</a>, <a href="https://registry.coder.com/modules/coder/agentapi">AgentAPI</a>, <a href="https://registry.coder.com/modules/coder/local-windows-rdp">Local RDP Button</a></li>
</ul>

These updates make it easier for teams to build with and contribute to Coder in a scalable, governed way.

From quick-start cloud templates to plug-and-play integrations, the Registry is how innovation gets shared across the Coder ecosystem. Whether you&#39;re building for your team or the community, the path to contribution is now open, supported, and scalable

<h4>Contribute and win a cool prize ✨</h4>

To celebrate this launch, we’re kicking off a Registry Bounty Program and we want your contributions.

Between now and August 15th, submit a module or template that matches one of our <a href="https://github.com/coder/registry/issues?q=is%3Aopen%20is%3Aissue%20label%3Abounty">in-demand contribution ideas</a> and claim the GitHub issue linked to the bounty. If your contribution is accepted, you’ll earn exclusive prizes and a shoutout in our community.

Whether it’s a module for an AI agent, a workspace helper, or an integration with a popular dev tool, we’re looking for contributions that help teams go from zero to productive — fast.

<h3>Get started with the Coder Registry</h3>

<a href="https://registry.coder.com/">Browse the Registry</a>, and check out our new <a href="https://github.com/coder/registry/blob/main/CONTRIBUTING.md">contribution guidelines</a> to submit your own modules. 

<hr>

<h2>ICYMI: July 2025 Launch Week lookback 👀</h2>

What a week. Over five days, we highlighted powerful new features that make Coder faster, more secure, more flexible—and more ready than ever for the future of development. 

Here&#39;s a quick recap of everything we launched:

<h3>🧠 Day 1: Coder Tasks – AI agents you actually control</h3>

AI coding agents are here. Most tools force you to choose between security and usability.

<a href="https://coder.com/blog/launch-week-2025-introducing-coder-tasks">What we shipped</a>:

<ul>
<li>Coder Tasks: Run Claude Code, Amazon Q, and other agents in secure, self-hosted environments with real-time control.</li>
<li>Isolated agent workspaces, GitHub integration, and a fully vendor-neutral architecture built for scale.</li>
</ul>

✨ Agentic automation, without cloud lock-in.

<h3>🛡️ Day 2: Intelligent governance – Guardrails you don’t hate</h3>

Governance often comes at the cost of developer freedom. We don’t think it has to.

<a href="https://coder.com/blog/launch-week-2025-intelligent-governance">What we shipped</a>:

<ul>
<li>Dynamic Parameters: Smart forms that flex with developer input. No duplicated templates or fragile workarounds.</li>
<li>Agent Boundaries: Dual-layer isolation to safely run agents and humans in the same workspace.</li>
<li>Workspace Presets: Reduce onboarding friction with easy workspace creation in just a few clicks.</li>
<li>Notifications: Keep teams in the loop on important events.</li>
</ul>

✨ Governance that scales and stays out of your way.

<h3>🖥️ Day 3: Hybrid development – Remote dev that feels local</h3>

Remote development only works if developers actually want to use it.

<a href="https://coder.com/blog/launch-week-2025-hybrid-workflows">What we shipped</a>:

<ul>
<li>Coder Desktop: Instant, secure connectivity for a seamless development experience that just works.</li>
<li>MCP Server: First steps toward unified, native-feeling cloud/local development.</li>
</ul>

✨ Portability, meet performance. You’re going to get along just fine.

<h3>⚡ Day 4: Instant infrastructure – No setup required</h3>

Developers shouldn’t have to wait for their environment to catch up.

<a href="https://coder.com/blog/launch-week-2025-instant-infrastructure">What we shipped</a>:

<ul>
<li>Prebuilt Workspaces: Start coding instantly—on any branch, in any repo.</li>
<li>Dev Containers Rework: Full spec support, multi-container setups, automatic rebuilds.</li>
<li>JetBrains Toolbox Plugin: Official way to run and manage JetBrains IDEs in Coder.</li>
</ul>

✨ Developer environments, ready before you are.

<h3>🧩 Day 5: Composable by design – A smarter way to build</h3>

Every team needs something different. Coder’s curated marketplace of templates and modules makes it easy to create and share.

What we shipped: 

<ul>
<li>It&#39;s easier than ever to contribute</li>
<li>Governance is clearer and more scalable</li>
<li>The platform is more stable and performant</li>
</ul>

<h2>What’s next</h2>

Launch Week might be wrapping up, but we’re just getting started. This week marked a major step toward agent-native development — with Coder Tasks and smarter governance forming the foundation for bringing AI into real, production-grade workflows. Combined with instant workspaces, local-feeling remote dev, and a growing ecosystem of reusable modules, the path to scalable, secure software development has never been clearer.

We’ve been energized by the feedback, the questions, and the early adoption of what we’ve launched—and we’re especially excited to see how you push these new features further.

→ Have an idea? <a href="https://coder.com/docs/install?utm_campaign=25Q2PD_H1_Product_Launch_Week&utm_source=coder&utm_medium=email&utm_content=product&utm_term=">Install Coder open source</a> and start building.

→ Want to share feedback and connect with others? <a href="https://discord.gg/coder">Join our Discord</a>.

→ Not sure what to do next? <a href="https://coder.com/contact/sales">Talk to our team</a> — we’re here to help.

→ Looking for more future-looking insights? <a href="https://coder.com/events-webinars/how-anthropic-engineers-are-adapting-to-ai-coding-agents/register">Grab a seat</a> to our upcoming webinar with the Developer Experience team at Anthropic, the company behind Claude Code, where we&#39;ll discuss the impact of AI agents on software development.

The Coder Registry is a growing marketplace for secure, enterprise-ready modules — built to share, remix, and accelerate innovation across teams.

Day 5: Composable by Design – What will you build? + Launch Week Recap

This post is part of Coder Launch Week (July 7–11, 2025). Each day, we’re sharing innovations that make secure, scalable cloud development easier. Follow along <a href="https://launch.coder.com">here</a>. 

With a desktop app for connecting to remote workspaces, an integration surface for external agents, and one-click JetBrains IDE launches, Coder is making remote development feel local—and giving developers and autonomous tools more control over how environments are created and used.

<h2>Coder Desktop: Bridging the gap between local and remote development</h2>

Remote development has come a long way, but it&#39;s still never felt local. Until now.

Coder Desktop Demo | Coder Connect

<h3>Challenge: Remote development promises speed, but there are still growing pains</h3>

Remote development unlocks computing resources far beyond what most local machines can offer — high-performance GPUs for machine learning and graphics processing, specialized CPU architectures for demanding workloads, and HPC clusters for massive parallel processing. But despite this raw power and flexibility, remote development often feels clunky, fragmented, and fragile compared to the speed and responsiveness of local setups. 

That’s where Coder comes in.

Coder is an open platform that provisions secure, cloud-based development environments on your infrastructure. It centralizes and automates workspace creation, offering a consistent, scalable way to develop in the cloud. Traditionally, developers interact with Coder workspaces using the <a href="https://coder.com/docs/install/cli">Coder CLI</a>, a command-line interface that connects local tools to remote environments. But even with these benefits, developers who are used to their local machines sometimes find the transition difficult.

Here’s why:

<ul>
<li>Developers must manually install and maintain the Coder CLI to ensure version compatibility with their Coder deployment</li>
<li>Tool access is tedious and error-prone, requiring commands like <code>coder port-forward</code> for each new service (e.g., devtools, web apps, databases)</li>
<li>Losing <code>localhost</code> access breaks the feedback loop, disrupting WebSocket connections, hot reload, and browser extensions</li>
<li>Remote setups restrict hybrid workflows when GUI apps or analysis tools need direct access to remote files—or when you need to work offline</li>
<li>Special configuration or SSH tricks are often needed to connect GUI apps and terminal-based IDEs to remote workspaces, limiting flexibility</li>
<li>Non-VS Code IDEs require manual setup without native integration; tools like <a href="https://registry.coder.com/modules/coder/zed">Zed</a> or <a href="https://registry.coder.com/modules/coder/jetbrains-fleet">Fleet</a> depend on custom SSH configs</li>
</ul>

These gaps create friction at every layer — onboarding, iteration, debugging, and tooling — and limit the flexibility modern developers need to use the best tool for the job.

<h3>Solution: Instant, secure connectivity with Coder Desktop</h3>

Coder Desktop eliminates these friction points by providing instant, secure connectivity without CLI installation, version management, or manual port forwarding. It handles all complexity behind the scenes, delivering a smooth development experience that just works.

<img src="https://www.datocms-assets.com/19109/1752037192-coder-desktop-1-edited.png" style="max-height: 1000px; width: auto;" />

Coder Desktop includes three game-changing features:

<ul>
<li>Coder Connect: Access your remote workspace and resources without extra steps—no port forwarding required</li>
<li>Native IDEs and tools: Launch <a href="https://registry.coder.com/modules/coder/zed">Zed</a>, <a href="https://registry.coder.com/modules/coder/jetbrains-fleet">Fleet</a>, and full GUI desktops with a single click through RDP and VNC clients—no Coder CLI needed</li>
<li>File sync: Automatically synchronize your local and remote files in real time–no manual intervention</li>
</ul>

Coder Desktop combines the best of both worlds: the scalability, consistency, and security of cloud development and the convenience, speed, and familiarity of local development. 

<ul>
<li>For developers: Enjoy fast, local-feeling workflows with your preferred tools—no X11 forwarding, browser lock-in, or setup friction</li>
<li>For team leads and IT decision-makers: Onboard securely, standardize environments, and enforce access policies without limiting developer flexibility</li>
</ul>

<h4>Coder Connect: Local-feeling remote access</h4>

Coder Connect brings the convenience of <code>localhost</code> back to remote development. With a single toggle, each workspace gets a <code>.coder</code> hostname (like <code>myworkspace.coder</code>), making remote services feel like they’re running on your local machine.

This opens up powerful hybrid workflows. For example, you can run your frontend React app locally while connecting to a powerful remote backend in your Coder workspace. Just point your dev server to <code>http://myworkspace.coder:8080/api</code>–no CORS issues, proxy config, SSH tunnels, or port-forwarding rules.

All your remote workspace services become available through the same hostname:

<ul>
<li><code>http://myworkspace.coder:3000</code> → remote Node.js app</li>
<li><code>myworkspace.coder:5432</code> → PostgreSQL (via pgAdmin)</li>
<li><code>myworkspace.coder:27017</code> → MongoDB (via Compass)</li>
</ul>

This unified model makes remote development feel cohesive and familiar. It’s especially valuable for modern web stacks: your Vite dev server can run remotely while browser extensions, dev tools, and hot reload behave exactly as they would locally. WebSockets and Server-Sent Events remain stable and responsive. 

Coder Connect uses <a href="https://www.wireguard.com/">WireGuard</a>® to establish a fast, encrypted tunnel between your machine and the workspace. The connection is secure and low-latency: reliable enough for real-time workflows, and resilient across networks or mobile hotspots.

Whether you&#39;re running everything remotely or blending local and remote services, Coder Connect handles the networking complexity for you. It delivers the most native-like remote development experience available, letting you focus on building great web applications rather than managing infrastructure.

Getting started is incredibly simple. Just install Coder Desktop, enable Coder Connect, and you&#39;re ready to code.

<h4>File sync: Real-time hybrid workflows</h4>

Remote development excels at offloading resource-intensive compute and centralizing code. But what about cases where you need files locally–to use a GUI graphics editor app, run code analysis tools, or work offline on a plane?

Coder Desktop’s file sync solves this. Powered by the high-performance <a href="https://mutagen.io/">Mutagen</a> engine, it provides real-time, bidirectional synchronization between your local folder and remote Coder Workspace. It acts like a live mirror: changes on either side propagate almost instantly to the other. 

Mutagen was purpose-built for bridging local and remote development. It uses low-level filesystem monitoring to detect changes immediately, transmits only file deltas (not full files), and handles large codebases and binary assets efficiently. Compression and parallel transfer keep syncing fast even over standard broadband. The result is near-instant code propagation without network congestion—and no configuration complexity.

This enables key hybrid development use cases:

<ul>
<li>Use local GUI apps with remote projects: Open design files, images, or datasets in native tools like Photoshop or JupyterLab while letting compute-heavy tasks run remotely.</li>
<li>Accelerate search and code navigation with local tools: Use grep, <code>find</code>, or local IDE indexing for faster code inspection and editing—ideal for large repositories or nested file structures. Your edits stay in sync with the remote workspace without extra steps.</li>
<li>Offline development: Working during travel or internet outages? Maintain a local copy with file sync, and changes made offline will be reconciled automatically when you reconnect.</li>
</ul>

Think of it like an intelligent <code>rsync</code>—syncing files directly and securely between your local machine and remote workspace, with no cloud storage in the middle and no manual setup required.

Setup is simple. Open the File Sync panel in Coder Desktop, choose a local directory and a remote path, and click Add. Syncing begins immediately and continues in the background, automatically tracking file creations, edits, and deletions on both sides.

<img src="https://www.datocms-assets.com/19109/1752035369-file-sync-1.png" style="max-width:100%" />

<h4>Custom URI schemes: One-Click IDE Connections</h4>

Modern development requires flexibility in tool choice. While Coder Desktop offers native integration for VS Code through extensions, many developers rely on other IDEs, GUI environments, or the terminal. Coder’s new URI schemes and deep linking make it easy to launch any application and connect it to your remote workspace. Connections are encrypted, authenticated, and tool-agnostic—so you can use whatever workflow suits you best.

No lock-in. No workarounds. Just open and connect.

One-click GUI access via RDP

Coder supports Remote Desktop Protocol (RDP) deep links. When your workspace template includes an RDP setup, Coder exposes a <code>coder://</code> link that securely handles workspace authentication and addressing. One click launches your native RDP client directly into the workspace&#39;s GUI environment. 

Here is a <a href="https://registry.coder.com/modules/coder/local-windows-rdp">module</a> from Coder Registry to add this functionality to your workspaces.

<pre><code class="hcl">module &quot;rdp_desktop&quot; {
 count = data.coder_workspace.me.start_count
 source = &quot;registry.coder.com/coder/local-windows-rdp/coder&quot;
 version = &quot;~&gt; 1.0&quot;
 agent_id = coder_agent.main.id
 agent_name = coder_agent.main.name
}
</code></pre>

This creates a one-click “Open in RDP” button in the workspace dashboard and Coder Desktop. No manual setup, credential copying, or tunnel configuration.

<img src="https://www.datocms-assets.com/19109/1752008614-button-rdp.gif" style="max-width:100% />

Native SSH access with no extra config

For IDEs like <a href="https://registry.coder.com/modules/coder/zed">Zed</a> or <a href="https://registry.coder.com/modules/coder/jetbrains-fleet">Fleet</a> without native Coder integrations, Coder Connect enables native SSH without needing the Coder CLI or custom SSH configuration. Just run:

<pre><code class="bash">ssh myworkspace.coder
</code></pre>

Automatic DNS resolution and port forwarding eliminate the need for complex connection strings or manual setup.

<h3>Try Coder Desktop</h3>

Ready to simplify remote development? <a href="https://coder.com/docs/user-guides/desktop">Download Coder Desktop</a> for macOS and Windows (requires Coder v2.20+) and start exploring these powerful features today.

Join our community <a href="http://discord.gg/coder">Discord</a> or <a href="https://github.com/coder">GitHub</a> to share your feedback and experience. We can&#39;t wait to see your new use cases!

<hr>

<h2>MCP Server: Enable AI agents to orchestrate Coder</h2>

Coder now supports the <a href="https://modelcontextprotocol.io/introduction">Model Context Protocol</a> (MCP), so AI coding agents can create, configure, and manage dev environments with simple commands.

<h3>Challenge: AI agents are locked out of the toolchain</h3>

Earlier this week, we introduced <a href="https://coder.com/docs/ai-coder">Coder Tasks</a>—a way to run coding agents like Claude Code, Aider, and Goose inside Coder workspaces.

But many teams also use external agents—those running in chat clients, local apps, or separate services. To be effective, these agents need a way to create, manage, and interact with cloud workspaces as part of broader, multi-step workflows.

<h3>Solution: Expose Coder through a protocol interface</h3>

The MCP server enables these external agents to connect to Coder workspaces through a lightweight, open protocol. They can provision environments, execute code, and interact with templates—without needing to run inside the workspace.

While Coder Tasks supports internal execution, the MCP server opens up external orchestration—so agents can treat Coder like part of their environment. These agents can perform actions such as:

<ul>
<li>Creating workspaces</li>
<li>Running commands on workspaces</li>
<li>Listing, starting, and stopping workspaces</li>
<li>Creating and updating templates (admin-only)</li>
</ul>

Our goal is to integrate, not dominate. We support modern and open infrastructure to build with all the tools you love–and the ones you’ll use next.

The MCP server runs locally via the Coder CLI, which includes helper commands for configuring tools such as Claude Code, Claude Desktop, and Cursor. This enables workflows like:

<ul>
<li>Writing code using agents from existing chat clients like Claude or ChatGPT, using Coder workspaces as the backend</li>
<li>Developing locally with Cursor, while spinning up Coder workspaces through Cursor Chat via MCP</li>
</ul>

We’re currently working on a remote MCP server + OAuth integration so serverless agents can connect without any local dependencies needed.

<h3>Try the Coder MCP Server</h3>

The MCP server is in beta. Read the MCP Server docs <a href="https://coder.com/docs/ai-coder/mcp-server">here</a>, and join the <a href="http://discord.gg/coder">Discord</a> to ask questions or share feedback.

<hr>

<h2>JetBrains Toolbox plugin: Cloud-ready IDEs in Coder workspaces</h2>

Run JetBrains IDEs in the cloud—without sacrificing performance or control.

<h3>Challenge: Bringing heavyweight IDEs to the cloud</h3>

JetBrains IDEs like IntelliJ IDEA, PyCharm, and CLion are trusted by developers working on complex, large-scale projects, from backend services to embedded systems. These tools offer rich language support, smart automation, and deep debugging capabilities that teams rely on every day.

But running them locally can introduce friction: managing updates across machines, dealing with performance variability, and provisioning compute-heavy projects on underpowered laptops.

<h3>Solution: JetBrains Toolbox integration for Coder</h3>

We&#39;re excited to announce official support for JetBrains Toolbox in Coder workspaces. 

JetBrains Toolbox replaces Gateway as the new standard for installing, updating, and managing JetBrains IDEs—all from a single, unified UI. Paired with Coder’s remote workspace platform, it gives teams the flexibility to run their preferred JetBrains tools on powerful cloud infrastructure—while maintaining a direct, near-local development experience.

<img src="https://www.datocms-assets.com/19109/1752037446-jetbrains-toolbox-in-coder-edited.png" style="max-height: 1000px; width: auto;" />

Benefits include:

<ul>
<li>Faster onboarding with centralized IDE provisioning</li>
<li>Version control and rollback through Toolbox UI</li>
<li>Improved performance in cloud-hosted environments and reduced laptop load</li>
<li>Platform team visibility into compute usage and IDE configurations</li>
</ul>

<h2>Get started with JetBrains Toolbox on Coder</h2>

If you are on Coder 2.24+ and Toolbox 2.7+, add JetBrains IDE buttons to your workspace page using the JetBrains <a href="https://registry.coder.com/modules/coder/jetbrains">module</a> from the Coder Registry.

<pre><code class="hcl">module &quot;jetbrains&quot; {
 count = data.coder_workspace.me.start_count
 source = &quot;registry.coder.com/coder/jetbrains/coder&quot;
 version = &quot;1.0.0&quot;
 agent_id = coder_agent.example.id
 folder = &quot;/home/coder/project&quot;
}
</code></pre>

For full setup and troubleshooting instructions, check our <a href="https://coder.com/docs/user-guides/workspace-access/jetbrains/toolbox">documentation</a> and the <a href="https://github.com/coder/coder-jetbrains-toolbox">plugin readme</a>. Learn more about the Toolbox app in <a href="https://www.jetbrains.com/help/toolbox-app">JetBrains’ documentation</a>.

Coder is making development more flexible, powerful, and agent-ready, supporting workflows across any setup.

Day 3: Seamless Hybrid Workflows – Local Freedom, Remote Power

This post is part of Coder Launch Week (July 7–11, 2025). Each day, we’re sharing innovations that make secure, scalable cloud development easier. Follow along <a href="https://launch.coder.com">here</a>.

We’re showcasing four governance features that prove developer freedom and enterprise control aren’t opposites—they’re better together. From dynamic config forms to agent-specific safeguards, Coder makes governance not just manageable, but intuitive.

<h2>Dynamic Parameters: Identity-aware config management for scalable cloud development</h2>

Powerful, intelligent forms that flex with user input

Dynamic Parameters Demo | Context-aware workspace build forms

Coder’s mission is to empower every developer with powerful, secure workspaces without any setup pains, security blind spots, or configuration drift. Our platform standardizes remote development environments so teams can deliver consistent, scalable, and governed developer infrastructure. 

But they have faced an unfair trade-off: adopt a standard template for security and control—but surrender the plugins, frameworks, and tools that actually make developers productive.

Today, we’re removing that compromise. Dynamic Parameters let platform teams publish a catalog of tools and resources—then gate, reveal, or hide each option with identity-aware rules. 

<ul>
<li>Developers get a personally-tailored ‘Create Workspace’ form.</li>
<li>Security teams get guaranteed guardrails.</li>
<li>Operators maintain just a handful of templates.</li>
</ul>

<h3>Challenge: Scaling environments for each persona</h3>

In a traditional setup, platform engineers might create multiple docker images, YAML templates, or Helm charts for every environment or user scenario. Maintaining numerous templates is cumbersome for engineering teams. Teams lacking a central and standardized approach risk configuration drift, forgotten updates, and unclear audit trails of who changed what and when. 

<img src="https://www.datocms-assets.com/19109/1751974733-core-developer-tools.png" style="max-width:100%" />

Cloud development platforms like Coder centralize environment configuration to solve these challenges. A small team of platform administrators maintains <a href="https://coder.com/docs/admin/templates">templates</a> and RBAC to deliver secure and powerful environments across the organization. 

For platform administrators, scale becomes the challenge. Supporting dozens of teams often means maintaining dozens of templates — each defining the environment for a specific team, product, or workload. Every new developer tool or use case generates support tickets for subject matter experts, who must update and maintain these templates across the board.

This creates a heavy maintenance burden, causing delays and degrading the developer experience. At the same time, administrators want to minimize configuration strain on end-users by avoiding overwhelming them with options.

How can teams maintain just a few templates while still supporting many personas?

Mature platform teams look to build what’s often called a developer catalog: a central library of approved, reusable configurations and environment definitions. This internal marketplace of pre-approved tools gives developers the control and flexibility they want, without threatening security restraints. 

<img src="https://www.datocms-assets.com/19109/1751974802-restricted-privilege.png" style="max-width:100%" />

<h3>Solution: One template, many safe paths</h3>

Dynamic parameters deliver this vision of a self-service developer catalog with built-in guardrails by presenting users with reactive, identity aware controls during the environment creation process. 

To use Dynamic Parameters, platform teams:

<ul>
<li>Define a small set of standard templates.</li>
<li>Declare every option – IDEs, GPUs, namespaces, frameworks, agents, and more – in Terraform right beside your provisioning logic.</li>
<li>Automatically enforce user identity, choices, and policies honoring Coder’s RBAC.</li>
</ul>

Platform teams maintain fewer templates, developers enjoy a richer catalog, and security sees every workspace born with policy baked in. This results in fewer templates, faster tooling rollout, and happier developers.

With Dynamic Parameters, standardization doesn’t mean sacrificing flexibility. Instead of wiring separate ACLs, you can declaratively control what options are visible or available to whom. For example, a single template can: 

<ul>
<li>Reveal GPUs to <code>data-science</code> roles</li>
<li>Hide prod namespaces from interns</li>
<li>Surface experimental AI agents only to power users</li>
</ul>

Every control supports Terraform functions and <code>count</code>/<code>for_each</code>, so forms react instantly to user choices. To ship a new tool: edit one file, run <code>coder templates push</code>, and your new tool appears in the workspace creation flow for every developer.

How does this relate to developer portals like Backstage?
Dynamic Parameters focus on secure, identity-aware workspace provisioning — dynamically adjusting config options at environment creation time.

This complements internal developer portals like <a href="https://backstage.io/docs/features/software-catalog/">Backstage</a>, which organize services, documentation, and tooling at a higher level. Many teams use solutions like Backstage as the portal front door, while relying on tools like Coder + Dynamic Parameters to power the secure, flexible workspace experience beneath.

<h3>Using Dynamic Parameters</h3>

How to apply RBAC-aware parameters

The <code>coder_workspace_owner</code> data source lets you conditionally expose parameters to users depending on which Coder groups they belong to. For example, only expose the <code>attach_gpu</code> parameter to those with the <code>admin</code> or <code>data-science</code> role:

<pre><code class="hcl">locals {
 needs_gpu = contains(data.coder_workspace_owner.me.groups, &quot;admin&quot;)
 gpus = [
 &quot;None (Default for non-admins)&quot;,
 &quot;NVIDIA GeForce RTX 4090&quot;,
 &quot;NVIDIA A100&quot;,
 ]
}

data &quot;coder_workspace_owner&quot; &quot;me&quot; {}

data &quot;coder_parameter&quot; &quot;attach_gpu&quot; {
 # conditionally attach GPU based on identity
 count = local.needs_gpu ? 1 : 0

 name = &quot;ide_selector&quot;
 description = &quot;Select a GPU for this workspace.&quot;
 mutable = true
 display_name = &quot;Attach GPU&quot;
 default = local.gpus[0]
 form_type = &quot;dropdown&quot;
 type = &quot;string&quot;


 dynamic &quot;option&quot; {
 for_each = local.gpus
 content {
 name = option.value
 value = option.value
 }
 }
}

</code></pre>

👉 <a href="https://playground.coder.app/parameters/br1TlGHvwF">Try it on the playground ↗️.</a>

You can also use groups to target a Kubernetes namespace for your workspace. To do this, associate namespaces with Coder groups, and let users select from those they’re authorized to access. 

<pre><code class="hcl">locals {
 groups = data.coder_workspace_owner.me.groups
}

data &quot;coder_workspace_owner&quot; &quot;me&quot; {}

data &quot;coder_parameter&quot; &quot;k8s_namespace&quot; {
 type = &quot;string&quot;
 name = &quot;k8s_namespace&quot;
 display_name = &quot;Target namespace&quot;
 description = &quot;Select a namespace for this workspace&quot;
 mutable = true

 dynamic &quot;option&quot; {
 for_each = local.groups
 content {
 name = option.value
 value = option.value # Apply a (pre/su)ffix with native Terraform
 }
 }
}

</code></pre>

👉 <a href="https://playground.coder.app/parameters/qQthjTs4yX">Try it on the playground ↗️.</a>

Using Terraform arguments for conditional forms

Conditional parameters respond to user input in the workspace creation flow, enabling complex branching forms which guide developers naturally through their options. Dynamic Parameters support almost all native Terraform syntax, so programming your devtool catalog directly integrates with startup scripts and provisioning. 

Here, we can show certain workspace configuration options based on the user’s previous selection and even limit their options based on the project they want to work on. 

<pre><code class="hcl">locals {
 coder_git_repos = [
 &quot;coder/coder&quot;, &quot;coder/code-server&quot;, &quot;coder/weno&quot;, &quot;coder/preview&quot;
 ]

 coder_ml_git_repos = [
 &quot;coder/ml-nexus&quot;, &quot;coder/models&quot;
 ]

 ml_framework_options = [
 &quot;None&quot;, # Some users may install their own
 &quot;PyTorch&quot;, &quot;TensorFlow&quot;, &quot;JAX&quot;
 ]
}

# Repo Picker
data &quot;coder_parameter&quot; &quot;git_repo&quot; {
 name = &quot;Git Repo&quot;
 description = &quot;Select a machine learning repository to see ML framework options. (I.E. coder/models)&quot;
 type = &quot;string&quot;
 form_type = &quot;dropdown&quot;
 default = &quot;coder/models&quot;

 dynamic &quot;option&quot; {
 for_each = concat(local.coder_git_repos, local.coder_ml_git_repos)
 content {
 name = option.value
 value = option.value
 }
 }
}

data &quot;coder_parameter&quot; &quot;ml_framework&quot; {
 # Show this parameter iff git_repo is one of the ML repos
 count = contains(
 local.coder_ml_git_repos,
 try(data.coder_parameter.git_repo.value, &quot;&quot;)
 ) ? 1 : 0

 name = &quot;ML Framework&quot;
 description = &quot;Select the primary ML framework for this project.&quot;
 type = &quot;string&quot;
 form_type = &quot;dropdown&quot;

 dynamic &quot;option&quot; {
 for_each = local.ml_framework_options
 content {
 value = replace(lower(option.value), &quot; &quot;, &quot;&quot;) # e.g. &quot;pytorch&quot;
 name = option.value # e.g. &quot;PyTorch&quot;
 }
 }
}

# Now, the ML framework selected by users working on ML projects
# can be referenced during provisioning with:
#
# data.coder_parameter.ml_framework.value
</code></pre>

👉 <a href="https://playground.coder.app/parameters/sij5zM5mTg">Try it on the playground ↗️.</a>

See <a href="https://coder.com/docs/admin/templates/extending-templates/dynamic-parameters">our documentation</a> for more input types and ways to leverage conditional parameters using Terraform syntax.

<h3>Get started with Dynamic Parameters</h3>

You can try out the new form-building capabilities and prototype a self-service developer catalog using the <a href="https://playground.coder.app/parameters">Parameters Playground</a>. You can <a href="https://coder.com/docs/admin/templates/extending-templates/dynamic-parameters">opt in</a> to Dynamic Parameters on a per-template basis as of v2.24.0, so upgrade and try the new system today! 

We’re excited to see what users create with this form logic. Share your Parameter Playground link on our <a href="https://github.com/coder">GitHub</a> or <a href="http://discord.gg/coder">Discord</a>!

<hr>

<h2>Agent Boundaries: Double-firewall workspaces for AI coding agents</h2>

<h3>Challenge: Coding agents are a new frontier that demands new guardrails</h3>

AI coding agents are transforming development workflows, but they introduce governance challenges that are still uncharted territory. Unlike human developers, agents don’t intuitively understand what’s safe, appropriate, or permitted. Without the right controls, they may:

<ul>
<li>Access tools, code, or systems they shouldn’t</li>
<li>Trigger destructive actions unintentionally</li>
<li>Create unmanageable overhead for platform teams trying to mitigate risks</li>
</ul>

Today’s workarounds—prompt engineering, constant authorization gates, rigid sandboxes—don’t scale and erode productivity.

According to the <a href="https://www.ibm.com/thought-leadership/institute-business-value/en-us/c-suite-study/ceo">IBM 2025 CEO Study</a>, 82% of enterprises piloting AI agents cite governance as the top blocker to full-scale deployment. What enterprises need now is a practical, scalable way to safely deploy agents without smothering innovation or adding complexity that leads to fatigue and human error.

Agent Boundaries from Coder provide exactly that: a purpose-built solution that makes safe agent integration as easy and scalable as managing your existing dev environments.

<h3>Solution: Purpose-built isolation that works</h3>

Agent Boundaries extend Coder’s secure workspace model—already hardened by isolation, fire-walling, and ephemeral environments—with an agent-specific layer of protection:

<ul>
<li>Process isolation: Agents run inside doubly firewalled environments with fewer privileges than human developers</li>
<li>Template-level safeguards: Define what tools, code, or systems agents can access, without brittle permission hacks or improvised sandboxing</li>
<li>Integrated design: Humans and agents safely cohabitate in production-grade environments without isolated sandboxes, performance tradeoffs, or loss of utility</li>
</ul>

<img src="https://www.datocms-assets.com/19109/1751975244-boundaries.png" style="max-width:100%" />

Think of it as “Docker for agents”: a clean, consistent, modular approach to containment that works across clouds, teams, and regions.

<h3>Get started with Agent Boundaries</h3>

👉 <a href="https://coder.com/contact">Get in touch with your account team for early access</a>

👉 <a href="https://discord.gg/coder">Join our Discord channel to get help and share feedback</a>

<hr>

<h2>Workspace Presets: Faster, safer, smarter workspace creation</h2>

As developers embrace agentic AI, speed and reliability in workspace provisioning become more important than ever. That’s why we’re excited to announce the GA of Workspace Presets—a new way to streamline the developer experience and reduce setup friction, especially in environments where workspaces are created frequently and at scale.

<h3>Challenge: Guesswork brings risk</h3>

Workspace templates offer a wide range of configurable parameters — CPU, RAM, disk size, OS, cloud region, and more. But that flexibility comes at a cost.

Users, especially new or non-technical users, may struggle to select the right options, introducing inefficiencies or even breaking downstream workflows. Defaults may not match project needs. And too much freedom leads to too much variance, making it harder for platform teams to manage resources, enforce standards, or debug issues when things go wrong.

<h3>Solution: Easy workspace creation in just a few clicks</h3>

Workspace Presets are predefined sets of workspace settings that template admins can configure and attach to templates. Developers simply select a preset when launching a new workspace. Selecting a preset auto-populates workspace parameters—saving time, preventing misconfigurations, and enabling consistency across teams.

Presets give admins the ability to curate proven, tested configurations, and developers the ability to create workspaces in just a few clicks.

Let’s take a look at how the presets are defined. Below is a code snippet from the template source code. As you can see, we’re defining two presets: one for frontend developers working with React, and another for backend developers working with Go.

<pre><code class="json">data &quot;coder_workspace_preset&quot; &quot;fe&quot; { 
 name = &quot;Frontend React team&quot; 
 parameters = {
 (data.coder_parameter.repo_base_dir.name) = &quot;coder/coder.com&quot;
 (data.coder_parameter.image_type.name) = &quot;codercom/oss-dogfood:latest&quot; 
 } 
}

data &quot;coder_workspace_preset&quot; &quot;be&quot; {
 name = &quot;Backend Golang team&quot;
 parameters = {
 (data.coder_parameter.repo_base_dir.name) = &quot;coder/coder&quot;
 (data.coder_parameter.image_type.name) = &quot;codercom/oss-dogfood-nix:latest&quot;
 }
}
</code></pre>

The selection of the preset from the list is very easy. Simply select a preset from the dropdown list and proceed with the workspace creation. Settings covered by the preset are now hidden but accessible.

<h3>Reliable infrastructure for agentic AI development</h3>

Agentic development is fast, iterative, and unpredictable. Workspaces may spin up automatically or in rapid succession to support experimentation, fine-tuning, or task decomposition.

Workspace Presets make that velocity manageable. They ensure that each workspace created adheres to a configuration that’s optimized, secure, and validated. By reducing the number of manual inputs and enforcing consistency, Presets help organizations adopt agentic AI without adding operational complexity.

As agentic AI reshapes how development gets done, Presets provide the foundation to scale it—safely, efficiently, and with confidence.

<h3>Get started with Workspace Presets</h3>

👉 <a href="https://coder.com/docs/admin/templates/extending-templates/parameters">Learn all the ways to extend templates</a>

👉 <a href="https://coder.com/docs/admin/templates/extending-templates/parameters#workspace-presets">View an example of preset parameters</a>.

<hr>

<h2>Notifications: Real-time alerts for fast-moving dev environments</h2>

As development environments are becoming more dynamic, event-driven, and autonomous, and users launch, modify, and terminate environments at scale, it becomes increasingly important to stay informed—instantly and reliably. That’s where Coder Notifications come in.

<h3>Real-time updates about key system events</h3>

Notifications in Coder were <a href="https://github.com/coder/coder/releases/tag/v2.21.0">first released as GA in April 2025</a>. They allow users and admins to receive real-time updates in response to key system events—such as a workspace being deleted, a user being created, or a system-level issue emerging.

These alerts can be delivered through multiple channels, including:

<ul>
<li>Email</li>
<li>Coder UI inbox</li>
<li>Webhooks (for integration with third-party systems)</li>
<li>Slack</li>
<li>Microsoft Teams</li>
</ul>

<h3>Bridging automated processes and human oversight</h3>

In dynamic organizations, environment creation, teardown, and task execution can happen more frequently. This increased system activity makes observability and reactivity more important than ever.

Coder Notifications provide a bridge between automated processes and human oversight. Whether it’s alerting an admin to a misconfigured workspace, or surfacing issues like resource contention or quota thresholds—notifications make the invisible visible, and actionable.

<h3>What’s next?</h3>

We have a series of upgrades planned for the second half of 2025:

<ul>
<li>Custom Notifications: Script alerts for long-running builds, GPU attachments, or maintenanc, keeping users informed beyond built-in triggers.</li>
<li>Multiple Notification Targets: Send alerts to multiple destinations for richer integrations with platforms like Slack and <a href="https://coder.com/blog/coder-dx-partner-developer-productivity-cloud-based-development">getDX</a>.</li>
<li>New Trigger Types: These include workspace ready, imminent auto-stop due to inactivity, quota nearing limit, and agent task completion and input prompts.</li>
</ul>

<h3>Get started with Notifications</h3>

👉 <a href="https://coder.com/docs/admin/monitoring/notifications">View the complete list of event types and configuration docs</a>

👉 <a href="https://coder.com/docs/admin/templates/extending-templates/resource-monitoring">Learn how to enable OOM/OOD notifications on a template</a>

We’re showcasing four governance features that prove developer freedom and enterprise control aren’t opposites—they’re better together. From dynamic config forms to agent-specific safeguards, Coder makes governance not just manageable, but intuitive.

Intelligent Governance – True Developer Self-Service

Day 2: Intelligent Governance – True Developer Self-Service

AUSTIN, TX – July 7, 2025 – Today, <a href="https://coder.com">Coder</a>, the AI software development company, introduces a major platform upgrade designed specifically for enterprise teams working with AI coding agents. This release brings secure, practical <a href="https://coder.com/solutions">AI-assisted coding</a> into everyday enterprise software development workflows, bridging the gap between promising AI technologies and real-world use.

The launch marks the next chapter in Coder’s mission to make agentic AI a safe, trusted, and integral part of every software development lifecycle. With today’s announcement, Coder is introducing AI cloud development environments (CDEs) that combine full-featured IDEs, dynamic policy governance, and seamless agent orchestration — all within enterprise-controlled infrastructure.

&quot;How engineers develop is growing and evolving quickly, especially as we introduce Claude Code internally,&quot; said Jacqueline Lee, Member of Technical Staff at Anthropic. &quot;With Coder&#39;s flexible architecture, we can seamlessly support developer environments for both humans and AI agents while keeping our engineers productive.&quot;

<h2>Removing the friction for AI in the enterprise</h2>

As AI transforms how code is written, tested, and shipped, enterprises face a daunting challenge: infrastructure built for humans isn’t built for agents. Agents need secure environments, granular permissions, fast boot times, and full toolchain access — all while maintaining governance and compliance.

“The future of software development is clear: human developers collaborating with autonomous coding agents,” said Rob Whiteley, CEO of Coder. “The latest frontier models have demonstrated the potential of agentic coding, while also revealing a key requirement: each agent needs its own secure development environment. Yet today’s solutions often lack the enterprise-grade security, auditability, and flexibility that organizations demand. With our latest release, Coder empowers enterprises to bring AI out of the lab and into production, ensuring velocity without compromising security or control.”

Coder’s upgraded platform is enhanced with new agentic AI and human productivity capabilities:

<ul>
<li>AI Workspaces: Fully isolated, ephemeral environments where AI agents and developers work side-by-side with full context and control.</li>
<li>Agent Boundaries: A dual-firewall model to precisely scope agent access, ensuring security without degrading productivity.</li>
<li>Coder Tasks: A unified interface for running and managing AI agents with full context, isolation, and control.</li>
<li>Prebuilt Workspaces: Development starts instantly on any branch, eliminating setup time for both agents and humans.</li>
<li>Coder Desktop: A local companion app that connects remote workspaces to native tools, enabling tasks that browser IDEs can&#39;t support, like desktop apps or offline work.</li>
<li>Dynamic Parameters: Intelligent workspace creation forms that flex to user input, enforce policy, and eliminate Terraform sprawl.</li>
</ul>

This release also introduces a redesigned <a href="https://registry.coder.com/">Coder Registry</a>, a marketplace of community and official modules and templates for sharing and managing secure, enterprise-ready dev environments, with built-in modules that accelerate time-to-value for popular AI agents like Claude Code.

<h2>Launch Week: July 7–11, 2025</h2>

To celebrate the release, Coder is hosting a virtual <a href="https://launch.coder.com">Launch Week</a> with daily product reveals, technical deep-dives, and demos. Register now at <a href="https://launch.coder.com">launch.coder.com</a>.

About Coder

Coder is the AI software development company leading the future of autonomous coding. Coder helps teams build fast, stay secure, and scale with control by combining AI coding agents and human developers in one trusted workspace. Coder’s award-winning self-hosted Cloud Development Environment (CDE) gives enterprises the power to govern, audit, and accelerate software development without trade-offs. Learn more at <a href="https://coder.com">coder.com</a>.

Coder’s latest release enables AI agents to become a practical and secure reality for enterprise software development teams

Coder Unveils Enterprise-Grade Platform for Self-Hosted AI Development Environments

AUSTIN, May 20, 2025 - Coder, the AI software development company leading the future of autonomous coding, has won a <a href="https://aidevsummit.co/awards/">2025 AI TechAward</a> in the AI Integration: Business-to-Business category. The AI TechAwards are the software industry&#39;s annual awards recognizing outstanding design, engineering, and innovation in developer technology across 20 categories.

&quot;We&#39;re thrilled to receive this recognition for our work in AI infrastructure, enabling enterprise and government developer teams to <a href="https://coder.com/solutions/ai-coding-agents">safely deploy AI coding agents at scale</a>,&quot; said Rob Whiteley, CEO of Coder. &quot;As AI becomes increasingly central to software development, Coder is committed to providing the infrastructure that allows enterprises to seamlessly integrate AI tools, agents, and workflows into their development environments. This award validates our approach and the value of AI context and boundaries to our customers.&quot;

Coder&#39;s platform enables organizations to provide consistent, secure, and powerful cloud-based development environments for human and agentic development teams. By centralizing development infrastructure, Coder allows enterprises to more easily adopt and scale AI-powered development tools, deploy autonomous coding agents, and improve developer productivity – all while maintaining security and compliance.

&quot;Developer tools and technologies are leading the way for software developers and engineers to build the global technology ecosystem. Coder&#39;s win showcases their leading role in the growth and innovation across the software industry,&quot; said Jonathan Pasky, Executive Producer of DevNetwork, producer of AI DevSummit and the 2025 AI TechAwards.

The AI TechAwards winners were selected from hundreds of nominees by the independent, expert-led DevNetwork Advisory Board. Criteria included technical innovation, industry attention and awareness, and adoption by the developer, engineering, and IT communities.

Coder will be presented with its 2025 AI TechAward during <a href="https://aidevsummit.co/">AI DevSummit 2025</a> (May 28-29 in San Francisco, CA and June 4-5 online), the leading global conference for AI engineering. 

To learn more about what Coder is doing with AI coding agents, <a href="https://coder.com/solutions/ai-coding-agents">see here</a>. Visit <a href="https://aidevsummit2025.sched.com/event/22Alb/sponsor-spotlight-coder-live-rafflegiveaway">DevSummit Booth 306</a> to learn more about Coder, and join Rob Whiteley’s OPEN Session on <a href="https://aidevsummit2025.sched.com/event/20Trk?iframe=no">AI Coding Agents: Three Keys to Successfully Scaling in the Enterprise</a>.

About Coder

Coder is an AI software development company leading the future of autonomous coding. Coder helps teams build fast, stay secure, and scale with control by combining AI coding agents and human developers in one trusted workspace. Coder’s award-winning self-hosted Cloud Development Environment (CDE) gives enterprises the power to govern, audit, and accelerate software development without trade-offs. Learn more at coder.com.

Contact:

<a href="mailto:Coder@pancomm.com">Coder@pancomm.com</a>

Coder, the AI software development company leading the future of autonomous coding, has won a 2025 AI TechAward in the AI Integration: Business-to-Business category.

2025 AI Tech Awards Coder

Coder Wins 2025 AI TechAward for Best Innovation in AI Integration: Business-to-Business

As soon as <a href="https://www.anthropic.com/news/claude-3-7-sonnet">Claude Code was released</a>, we were chomping at the bit to see what it could actually do. Not on toy problems, but on our real development work. So we started feeding it real GitHub issues from our open-source and internal repositories — issues created by both community members and Coder engineers — to see how well it could tackle actual engineering problems.

This post covers three of those experiments. For each one, we break down what happened, how close Claude Code got to a working solution, where it struggled, and what we learned. These aren’t cherry-picked success stories. We wanted to see the full range, from wins to failures to everything in between.

Claude Code handled some tasks impressively and stumbled on others, but it’s effective enough that some of our developers use it every day. What makes that more encouraging is that we only gave it GitHub issue links and a few small nudges. It had to explore the repo and figure out the rest on its own.

<h2>Task #1: New feature in an internal admin dashboard</h2>

Documented by:
Hugo, Growth Engineer at Coder. I contribute to our core product and internal tooling, doing both backend and frontend work.

Task:
We have an internal admin dashboard built in Next.js. It&#39;s about 20,000 lines of code. Among other things, it displays GitHub issues created by our customers. We wanted to make the issues sortable by when they were created and updated.

Outcome (TL;DR):
After prompting Claude Code with the issue title &quot;Allow sorting /customer-issues by creation/update time,&quot; it simply carried out the task. Without additional instructions, it identified the relevant files, added a toggle to the UI for selecting the sort preference, and implemented the sorting logic. I reviewed the code and asked in a single sentence to also allow choosing the sort direction. It implemented that too.

I noticed it had mistakenly flipped the meanings of &quot;ascending&quot; and &quot;descending,&quot; so I edited a single line of code to fix it. While testing the UI, I discovered a bug in the existing human-written implementation: it mutated an object containing the issue list on every React render, causing some text to duplicate after re-rendering. I described the bug as I saw it in the browser and mentioned that I suspected a mutated object was the cause. Claude Code found the faulty code and fixed it. Finally, I asked it to submit a pull request, and it did.

The result was 114 lines added and 28 deleted in a single file.

Learnings:

<ul>
<li>Coming into this task, I didn&#39;t expect Claude Code to be useful. My past experience with AI agents, like those in Cursor, was that their suggestions were unhelpful. But Claude Code made the correct changes right away with little direction.</li>
<li>Using Claude Code was faster than writing the code myself. I wasn&#39;t the original author of the code that needed to be changed, and Claude Code figured out what to do sooner than I could.</li>
<li>I believe its effectiveness came from a combination of the task&#39;s simplicity, the small codebase, and the use of off-the-shelf frameworks like Next.js and React.</li>
<li>Claude Code&#39;s success with this task convinced me to use it more. I&#39;ve found it fantastic for clear tasks in small codebases but less effective when the task is unclear or the codebase is large and complex.</li>
</ul>

Cost: Less than $5.

Time: 5 minutes of AI work, 40 minutes of me reviewing the code and verifying the changes were correct.

What we’d do differently: Not much — Claude Code did great.

<h2>Task #2: Native macOS development</h2>

Documented by:
Thomas Kosiewski, Senior Software Engineer. I work in the networking team of Coder in both the backend and the macOS native parts.

Task:
<a href="https://github.com/coder/coder-desktop-macos/pull/105">https://github.com/coder/coder-desktop-macos/pull/105</a>
When dealing with Apple’s entitlements and signing process, we were encountering issues with apparently macOS losing the stapled certificates or other unknown processes occurring in the background on the machine. There is an open issue on the Apple Forums about this happening to other folks years ago, but Apple moves in mysterious, yet important ways, to quote Severance here. Throwing the AI at this issue, is more of an attempt to potentially find something that might be hidden in some deep Apple docs or some obscure and dated APIs, that are not easily accessible anymore.

Outcome (TL;DR):
The AI managed to produce code after multiple back and forth that compiles and works on macOS. Somewhat remarkable given the many language changes Swift has seen over the years. As this was mostly systems related and required a bigger amount of human interaction with macOS (deactivating and removing a network extension), it was a 30/70 split between Claude code working and me verifying and prompting it. 

Learnings:

<ul>
<li>I was surprised that Claude managed to find the APIs required to remove a network extension from macOS. Those are not well documented and even hard for a human to Google. Also, generating code that works is somewhat of a feat itself. Being able to request moving some buttons or error texts from one section to another tab and back, without having to do it oneself, is very refreshing, and makes refactoring a bit less manual.</li>
<li>The code quality resulting from the Claude code in this case was not good. It was functional and great for a PoC, but implementing it into production-ready code will take some time. It sometimes omitted existing delegates and reimplemented already existing things. (It also claimed to implement something that it didn’t build.)</li>
</ul>

Cost: ~$10

Time: ~30 min API time, ~90 minutes total time

What we’d do differently:
Prompting Claude Code to write “production ready” code and providing it with enough context and memory to produce high quality code is yet to be researched and understood, in my opinion. Without prior knowledge and experience, being able to quickly add any feature or showcase an idea of how something might or should behave is amazing, getting it over the finish line is an entirely different challenge.

<h2>Task #3: Backend development in Go</h2>

Documented by:
Cian, Staff Engineer at Coder. Works primarily on backend tasks, familiar with container orchestration tools, and author of this particular part of the codebase. Has used AI tools casually in the past.

Task:
Modify a JSON REST API endpoint that invokes a predefined command to return a different part of the command output: <a href="https://github.com/coder/coder/pull/16866">https://github.com/coder/coder/pull/16866</a> 

Initial prompt: *&quot;The agent/agentcontainers package defines a DockerCLILister. Part of the functionality of this struct is to enable listing running containers. There is a bug in the existing implementation where the returned port in the WorkspaceAgentListContainersResponse is incorrect. It should return the port of the container accessible from outside the container, but currently returns the port inside the container.&quot; *

Outcome (TL;DR):
The agent partially solved the task, but required some re-prompting to arrive at this partial solution with explicit instructions to fetch data from NetworkSettings.Ports and to de-duplicate host ports mapped to the same container port (note: that there is some additional fine detail to this that wasn&#39;t originally apparent to me either).

It also needed to be explicitly prompted to add test coverage using the default settings – this could likely be improved by modifying the prompt settings – and had to be prompted to re-run all tests in the package after making changes.

I also noted a propensity of the agent to write overly complex code (see LHS of commit 077d5a) related to comparing expected versus actual test outputs.

Ultimately, I decided that there were insufficient underlying test data to ensure a correct solution, and elected to raise a separate PR with appropriate test fixtures. It is possible that this would have helped the agent in finding a correct solution.

Learnings:

<ul>
<li>The agent did less well than expected; it seemed to just attempt the bare minimum of a solution without much attention to correctness.</li>
<li>Tasks that require knowledge of external APIs and/or dependencies on external tooling seem to be difficult for this system to reason about. Providing example data up front and explicitly instructing the agent to address the provided test cases may help here, but also runs a risk of &#39;over-fitting&#39; to the test data. </li>
<li>The agent also seems to prefer smaller changes – for example, if changing a single line in a test case exposes a bug in the code being tested, it was observed to prefer changing the test code as opposed to the code under test. It had to be explicitly prompted not to modify the test code.</li>
</ul>

Cost: $2.65 

Time: ~1h wall time (included initial setup), 7m 34s API time

What we’d do differently:

<ol>
<li>Provide specific test cases and examples up-front, if possible – or instruct the agent how to generate valid test data.</li>
<li>Instruct the agent to run package-level tests after changes in each package to establish a positive feedback loop.</li>
<li>Enforce Test-Driven-Development (TDD) by providing specific instructions about which sets of files the agent may or may not modify. If there is a known bug, first instruct it to add a test case to reproduce the bug while disallowing edits of the code under test. Then instruct it to fix the code under test while disallowing edits of the test code.</li>
<li>Remind the agent to perform a final refactoring step at the end (the oft-overlooked part of TDD!).</li>
</ol>

<h2>Conclusion</h2>

These experiments gave us a clearer sense of what AI agents can and can’t do today. Claude Code did well on small, well-defined tasks in familiar frameworks like React and Next.js. It struggled when deeper reasoning, system complexity, or poorly documented APIs were involved. What made this possible was running AI in isolated Coder workspaces, giving it a safe place to experiment without putting real code at risk. And while AI still needs human supervision, some of us are already using it daily to get through parts of our work faster.

We also came away with a better sense of costs, which is critical if you’re thinking about using AI at scale. Knowing what kinds of tasks AI can handle, how much it costs, and how to control that work safely is where the value starts to show. AI isn’t magic, but when paired with the right environment, it can be useful and is improving quickly. If you’re exploring this space too, <a href="https://coder.com/chat">we’d love to hear what you’re learning</a>.

Next, we’re planning a follow-up where we explore how far Claude Code can go with better prompts, more context, and internal tooling to make our codebase more AI-friendly. Stay tuned for part two.

We gave Claude Code real GitHub issues from our open-source repo to see how well it could handle actual engineering tasks. This post breaks down what worked, what didn’t, and what we learned. Including real insights on costs and where AI agents might already be useful (and where they still struggle).

Solving GitHub Issues with Claude Code

Platform engineering has become a hot topic in tech circles, with many organizations struggling to implement it effectively. Whether you call it “platform engineering”, “site reliability engineering”, or “developer experience”, the mission is the same: designing and maintaining an organization’s software infrastructure so that developers have a stable, cost-effective foundation for their applications and their development.

Coder recently hosted roundtable discussions with platform engineers to explore their experiences. What began as technical deep dives quickly evolved into candid conversations about the organizational dynamics shaping their roles.

<h2>Platform Engineer’s Balancing Act</h2>

Platform engineers deal with challenging audiences:

<ul>
<li>Developers expect simplicity, stability, and performance. </li>
<li>Organizations expect a return on their software investment. </li>
<li>Both see this infrastructure as a black box that the platform engineering team is responsible for. </li>
</ul>

The engineers we spoke to shared how these expectations manifest in their daily work:

<ul>
<li>Buy-in from stakeholders: Navigating internal politics and winning executive support.</li>
<li>Encouraging adoption: Choosing an effective mix of carrot and stick incentives. </li>
<li>Managing expectations: Balancing the need for quick solutions with long-term sustainability.</li>
<li>Measuring ROI: Convincingly justifying technical decisions with measurable results. </li>
</ul>

<h2>The Case for Platforms</h2>

With robust platforms in place, developers are free to focus on delivering applications instead of wrestling with infrastructure. The industry has all but officially standardized on cloud-native infrastructure services like Kubernetes and Docker, but they’re just the beginning. Platforms provide essential abstractions, managing the complexities of billing, web, API, message queuing, dev tools, and so on. 

For example, Coder helps <a href="https://coder.com/success-stories">large enterprises</a> measurably, dramatically lighten the burden of administering complex development environments. Developers get coding quickly with shortened project onboarding and build times. Organizations spend less on development infrastructure. Platform teams benefit from consolidated management and monitoring of development environments, freeing them from one-off issues and maintenance headaches.

<h2>Technical challenges and the over-general swamp</h2>

While technical challenges are part of the job, they often spiral into something bigger:

<ul>
<li>Growing variation: With a growing number of applications to support, each uses slight variations in the infrastructure they depend on. Over time, complexity creeps in. The glue that connects all this infrastructure becomes increasingly expensive to maintain.</li>
<li>Keeping up: As applications grow in scope, the platforms under them must scale too. The platform team stays ahead of this by updating or replacing components in the infrastructure. </li>
</ul>

The solutions to these technical challenges are defined and executed by platform engineering teams, but developers and the organization push back because they perceive changes in infrastructure as putting stability and ROI at risk. 

A platform team finds itself in an &quot;over-general swamp&quot; as they spend too much time maintaining disparate systems instead of innovating, while the application teams they support are drowning in complexity.

The goal? Escape to a &quot;golden path&quot; that checks off all these boxes:

<ul>
<li>Standardized well-supported platforms for common development features</li>
<li>Reduced cognitive load for developers</li>
<li>Flexibility for edge cases</li>
<li>Follow opportunities for innovation</li>
</ul>

A new book published by O’Reilly Media, Platform Engineering: A Guide for Technical, Product, and People Leaders, shows a platform engineering team how to leave the over-general swamp to find their golden path.

It’s written by leaders in the industry with decades of combined experience. Camille Fournier is an executive, speaker, and author. Ian Nowland’s resume includes Senior Vice President of Core Engineering at Datadog and, most recently, founder of Junction Labs. 

The book guides a platform team on making competent technical choices that win acceptance from stakeholders:

<ul>
<li>Build trust: Earning confidence through operational excellence and stakeholder buy-in for major investments.</li>
<li>Metrics: Creating measurable goals to justify technical decisions.</li>
<li>Standardization and innovation: Reducing complexity without stifling experimentation. </li>
<li>Managing migrations: Using more automated, self-service approaches.</li>
</ul>

<h2>Learn more</h2>

Our roundtable discussions highlighted the intersection of technical expertise and organizational savvy in platform engineering. That’s why we’re excited to share a free excerpt from Platform Engineering: A Guide for Technical, Product, and People Leaders:

<ul>
<li>Download the excerpt <a href="https://coder.com/ebooks-and-reports/ebooks/o-reilly-platform-engineering-ebook">for free</a>. </li>
<li>Read the <a href="https://www.oreilly.com/library/view/platform-engineering/9781098153632">full book from O&#39;Reilly</a> or your favorite book seller. </li>
<li>Hear from the authors directly at this webinar hosted by O’Reilly—<a href="https://www.oreilly.com/live-events/in-conversation-camille-fournier-and-ian-nowland-on-platform-engineering/0642572012003/">register here</a>.</li>
</ul>

Platform engineering has become a hot topic in tech. We share key insights from two Coder-hosted roundtables and the O'Reilly eBook, Platform Engineering: A Guide for Technical, Product, and People Leaders.



AI-native Development

Secure Environments for 2,500+ Developers

Coder Team

Explore the latest from Coder

Coder Sets a New Standard for AI Coding with Self-Hosted, AI Model Agnostic Coder Agents

Coder Ranks 15th on the 2026 Inc. 5000 List of the Fastest-Growing Private Companies in the Southwest

Coder Secures $90M Series C Led by KKR to Advance Secure Enterprise AI Development

Coder Wins 2026 ChannelVision AI Award for Partner Relationship Innovation

Coder Achieves AWS Generative AI and DevOps Competencies

Coder Listed in AWS “ICMP” for the US Federal Government

Coder Introduces AI Development Infrastructure for Hybrid Human and Agent Teams

Coder Launches Blink: The Fastest Way to Build and Deploy AI Agents

Related content

Using Claude Code with Coder Workspaces

Introducing Coder Agents

Self-Hosted Doesn’t Always Mean What It Implies