
Announcing the general availability of AI Bridge and Agent Boundaries in Coder v2.30, both included in Coder’s new AI Governance Add-On.
Agentic AI is moving fast, and it’s already inside enterprise development workflows. As coding agents gain the ability to execute commands, access networks, and act autonomously, the risks grow just as quickly. Platform teams don’t need best‑effort guardrails: they need governance that’s built into the environment.
Today, we’re announcing the general availability of AI Bridge and Agent Boundaries, both included in our new AI Governance Add‑On for Coder Premium customers. Together, these capabilities turn Coder workspaces into a governed control plane for AI coding agents — where identity, observability, and enforcement are architectural, auditable, and invisible to developers.
AI Bridge is Coder’s LLM gateway for auditing AI sessions, managing MCP servers, and enforcing policy across every AI tool your developers use.
Since beta, AI Bridge has expanded in three major ways:
AI Bridge eliminates API key sprawl, centralizes audit trails for compliance, and makes AI usage measurable across teams — all while leveraging your existing identity and workflows.
Agent Boundaries brings process-level network controls and audit logging to AI agents running inside Coder workspaces.
What’s new in GA:
Landjail mode, which runs without requiring permission changes — making Boundaries usable in far more environments.Agent Boundaries treats AI agents as untrusted actors. Instead of relying on brittle prompts or unusable sandboxes, it enforces runtime policies that strictly define what agents can access or export. This approach helps mitigate risks like prompt injection, data exfiltration, and unintended destructive actions by enforcing network‑level controls, not best‑effort instructions.
Learn more about Agent Boundaries.
Starting with Coder 2.30, we’re introducing the AI Governance Add-On for Coder Premium customers, designed to help teams move AI agents from individual experiments to a shared, production-grade platform capability.
Most teams begin with AI agents running locally or per-user. That works until agents need access to real code, real infrastructure, or to run at scale. The AI Governance Add-On gives platform and security teams architectural, auditable controls over AI usage without disrupting developer workflows.
With the add-on, teams can:
The add-on includes:
Together, these capabilities move AI agents into Coder’s platform, where they can be observed, controlled, and scaled like any other critical part of the development lifecycle.
Learn more in the AI Governance Add-On documentation.
Coder 2.30 includes a performance improvement that significantly speeds up workspace starts for templates that use Terraform modules. Instead of downloading modules every time a workspace initializes, Coder now captures them once when a template version is created and reuses that exact set on every subsequent start.
The result is faster, more predictable workspace startup, especially for platform teams managing shared templates or teams with complex Terraform-based setup. Each template version consistently uses the same module code, so workspace behavior is easier to reason about and less prone to unexpected changes.
Breaking change to note: Previously, modules—particularly those pointing to a Git branch—could update automatically on workspace start. Now, module updates require creating or re-importing a new template version.
This update is part of our broader effort to continuously reduce workspace start times across Coder. Learn more on our GitHub.
Don’t miss Community Week (Feb 9–13), where we’ll share our community vision, spotlight real builders, and make it easier to get involved with Coder. We’re kicking it off with Community Office Hours on Feb 12 at 10:30am PST. Join the community on Discord to add it to your calendar.

Jiachen Jiang, Senior Product Manager at Coder, will walk through the infrastructure patterns that prevent shadow AI, reduce blast radius, and keep security tight—without slowing developers down. Join us for real-world examples, a live demo, and a practical blueprint for governing AI agents at scale.
Also join our Agentic Workshop on March 26 at 10am PST for a hands-on session exploring how AI coding agents actually work inside secure, production-like development environments. Register here–we'll see you there!
Want to stay up to date on all things Coder? Subscribe to our monthly newsletter and be the first to know when we release new things!