Coder 2.27

The October release is the latest mainline release, focusing on AI governance and integrations, while improving the user experience.

View the full changelog on GitHub.

AI Governance and Security

AI Bridge in Early Access

• AI becomes auditable: built-in LLM proxy for OpenAI and Anthropic for collecting & aggregating LLM usage
• Centralized key management: developers use their Coder key to access LLMs, all actions attributable to users
• Automatic cost attribution: token usage breakdown per user/group/organization
• Centralized MCP management: configure MCP servers centrally, audit all tool calls issued along with their parameters

Setup guide

Agent Boundaries in Early Access

• Process-level firewall for coding agents with policy-driven controls.
• Granular restrictions that limit what AI coding agents can access or do, without blocking automation or complicating workflows.
• Network rule enforcement to block domains, subnets, or HTTP verbs and prevent data exfiltration.

New Terminal-based AI Agent Support

• GitHub’s Copilot CLI is the latest addition, which can be run inside Coder workspaces.
• Add agents with just a few lines of Terraform from the Registry, and integrate them with Coder

Developer Experience

GitHub to Coder Tasks

• Plug-and-play integration setup for GitHub + Claude, which can be used as is or modified for other software components.

Custom Notifications

• Users now get alerts when long-running AI tasks finish or need input, keeping builder focus where it’s needed most.

Coder_app tooltips

• Descriptive tooltips in the coder_app to help your teams discover features faster and stay productive with contextual, in-product guidance.

Coder Tasks Notifications

• Users now get alerts only when long-running AI tasks finish or need input, keeping builder focus where it’s needed most.

View the full changelog on GitHub.