Skip to main content
The Namespace MCP server lets AI agents investigate builds and GitHub Actions jobs, analyze compute and CI usage, and manage Devboxes.

Connect your agent

Add the server to your agent, then log in.
Add the server for all your projects:
Log in by running /mcp in a Claude Code session, selecting namespace, and choosing Authenticate. You can also log in from the terminal:
To check the connection, run claude mcp list.

Choose access

The first time your agent connects, it opens Namespace in your browser. Review the requested permissions, then authorize the workspace it can access.
Authorize MCP access page in Namespace, showing the requested Read and Write permissions and a list of workspaces, each with an Authorize button
Read-only access only limits the Namespace MCP server. An agent with shell access can still use nsc or devbox with your local credentials.

Common use cases

Use MCP with local tools

For the most efficient workflows, install nsc, devbox, and the GitHub CLI alongside the Namespace MCP server. MCP connects the agent to your Namespace resources. When a task needs capabilities it doesn’t provide, it may guide the agent to the appropriate CLI.

Manage access

Each connection is authorized for one workspace. To use another workspace, add the server under a different name and authorize it separately. Sessions renew automatically. To review or revoke access, open Applications in the Namespace dashboard. Revoked agents must log in again.

Troubleshooting

Last modified on October 7, 2026