--- title: "Coder Agents - AI Coding Agents on Your Infrastructure | Coder" description: "Delegate development work to AI agents running on your governed, self-hosted infrastructure. Coder Agents provides centralized model governance, audit logging, cost controls, and enterprise-grade security with no API keys in workspaces." image: "/_next/static/media/OpenGraph.37da9236.jpg" canonical: "https://coder.com/solutions/agents" --- # A first-class agent experience on your infrastructure. Run Coder’s native agent on infrastructure you already control, from cloud VPCs to air-gapped environments. Give developers a first-class experience while keeping agent workspaces network isolated and tightly scoped. [Install Coder](https://coder.com/docs/install)[View documentation](https://coder.com/docs/ai-coder/agents) ![Two Column Masthead Background](https://coder.com/_next/static/media/twocol-premium-masthead-bg.8b7ffea4.jpg?w=3840)![Coder Agents chat interface](https://coder.com/_next/static/media/agents-nav-thumb.ea22fda4.png?w=3840) ## Where developers delegate coding work to agents and platform teams standardize the experience with enterprise-grade security, governance, and auditability. ![Background gradient](https://coder.com/_next/static/media/one-col-blue-bg.f50e775a.jpg?w=3840)![Coder Agents interface with parallel task execution](https://coder.com/_next/static/media/agents-hero-image.f39284a9.png?w=3840) For Platform TeamsFor Developers ### Deploy a standardized agent experience. The entire agent loop runs in the Coder control plane on infrastructure you control. Platform teams get centralized control over models, prompts, tools, and spend to deliver a consistent agent experience across the organization. - **Model Governance** — Centrally approve providers, models, private endpoints, and enterprise API keys so developers use only sanctioned options. - **Usage Limits & Cost Tracking** — Set per-user and per-group spend controls and monitor usage across models, users, and chats. - **No API Keys in Workspaces** — LLM credentials exist only in the control plane, eliminating a critical exfiltration path. - **Workspace Governance** — Restrict which workspace templates agents can use so they only run on approved infrastructure. - **User Identity on Every Action** — Tie every agent action to an authenticated identity through your existing SSO stack for attribution, auditability, and accountability. - **Standardized MCP and Skills** — Configure and extend agent behavior with centrally managed MCP tools, skills, and policies. ## A control plane for agents. Use a single control plane to run agents and automatically provision workspaces only when compute is needed. No separate service to deploy, and no extra agent platform to set up or manage. [![Cog icon](https://coder.com/_next/static/media/cog-white.f1180027.svg?w=3840)Workspace provisioning](https://coder.com/docs/ai-coder/agents#automatic-workspace-provisioning)[![Globe icon](https://coder.com/_next/static/media/globe-white.453239a2.svg?w=3840)Control plane architecture](https://coder.com/docs/ai-coder/agents/architecture#what-runs-where)[![Padlock icon](https://coder.com/_next/static/media/padlock-white.5b4e6a27.svg?w=3840)Security benefits](https://coder.com/docs/ai-coder/agents/architecture#security-posture) ![Coder Agents architecture with control plane and workspaces](https://coder.com/_next/static/media/key-white.84f7d782.svg?w=3840) ![Background gradient](https://coder.com/_next/static/media/one-col-red-bg.a4ff1e6b.jpg?w=3840)![Coder Agents model configuration interface](https://coder.com/_next/static/media/any-llm-provider.089ecf30.png?w=3840) ### Bring any model. Connect the models that work for your organization, from Anthropic, OpenAI, Google, and Bedrock to self-hosted endpoints. Switching models is a configuration change, not a platform migration. [![Box icon](https://coder.com/_next/static/media/box-white.5d9bc102.svg?w=3840)Model configuration](https://coder.com/docs/ai-coder/agents/models)[![Plug icon](https://coder.com/_next/static/media/plug-white.9962f38a.svg?w=3840)Supported providers](https://coder.com/docs/ai-coder/agents/models#providers)[![Claude icon](https://coder.com/_next/static/media/claude.487fd479.svg?w=3840)Anthropic](https://coder.com/docs/ai-coder/agents/models#anthropic)[![OpenAI icon](https://coder.com/_next/static/media/openai.eeed2506.svg?w=3840)OpenAI](https://coder.com/docs/ai-coder/agents/models#openai)[![Google Gemini icon](https://coder.com/_next/static/media/gemini.86c87b35.svg?w=3840)Google](https://coder.com/docs/ai-coder/agents/models#google)[![AWS icon](https://coder.com/_next/static/media/aws.dc0ca0fd.svg?w=3840)Bedrock](https://coder.com/docs/ai-coder/agents/models#configuring-aws-bedrock) ## Trigger agents anywhere. Use the chat UI for foreground work, or trigger background tasks through the Coder Agents API. Integrate with CI pipelines, GitHub Actions, Slack, Jira, and other systems to automate work end to end. [![Browser icon](https://coder.com/_next/static/media/browser-white.81827469.svg?w=3840)Agents API](https://coder.com/docs/ai-coder/agents/chats-api)[![Plug icon](https://coder.com/_next/static/media/plug-white.9962f38a.svg?w=3840)MCP integrations](https://coder.com/docs/ai-coder/agents/platform-controls/mcp-servers) ![Trigger Coder Agents from external systems](https://coder.com/_next/static/media/trigger-anywhere.b73d8c70.png?w=3840) ## Measure adoption and outcomes. Track usage, attribution, and outcomes across your organization, including token consumption, model selection, pull request activity, and merge rates. [![Log icon](https://coder.com/_next/static/media/log-white.518b38a8.svg?w=3840)Usage analytics](https://coder.com/docs/ai-coder/agents/platform-controls/usage-insights)[![Key icon](https://coder.com/_next/static/media/workspace-provisioning.a97e0457.png?w=3840)Cost tracking](https://coder.com/docs/ai-coder/agents/platform-controls/usage-insights#cost-tracking)[![Padlock icon](https://coder.com/_next/static/media/padlock-white.5b4e6a27.svg?w=3840)Usage limits](https://coder.com/docs/ai-coder/agents/platform-controls/usage-insights#usage-limits) ![Audit logs for Coder Agents activity](https://coder.com/_next/static/media/measure-outcomes.2e3cc5cc.png?w=3840) ### Self-hosted and fully inspectable. Run Coder Agents inside your VPC, on-prem environment, or air-gapped enclave. Source code stays inside your perimeter, and the agent architecture is open to inspection, extension, and control. [![Globe icon](https://coder.com/_next/static/media/globe-white.453239a2.svg?w=3840)Install Coder](https://coder.com/docs/install) ![Background gradient](https://coder.com/_next/static/media/one-col-red-bg.a4ff1e6b.jpg?w=3840)![Coder open source on GitHub](https://coder.com/_next/static/media/open-source-agent.9f2fbaa1.png?w=3840) ## FAQs What is Coder Agents? Coder Agents is a native AI coding agent built into Coder that runs on your self-hosted infrastructure, enabling developers to delegate work through chat or API while platform teams maintain control. How is Coder Agents different from other coding agents? Does Coder Agents require a separate deployment? Which LLM providers are supported? Can I control what agents have access to? How can I get started?