Skip to Content

Connect an external AI assistant (MCP)

Let Claude or another MCP-capable assistant create tasks, approve specs, start runs and deploy - with the access you approve.

If you already work with an AI assistant such as Claude, you can let it work with CodeLynq directly. CodeLynq speaks the Model Context Protocol (MCP): the assistant can list your projects, create tasks, take them through approval to a run, and follow the result - while you stay in the conversation you are already having.

The assistant acts as you. It can only do what your account can do, within the access you approve, and everything it does lands in the audit log under your name.

Connect an assistant

Add CodeLynq as an MCP server (a remote server over HTTP) in your assistant, with this address:

/api/v1/mcp

There are two ways to sign in:

  • Sign in and approve (OAuth) - for assistants that support it. The assistant opens a CodeLynq page that says {assistant} wants to access your CodeLynq account and lists what it will be allowed to do. If you are not signed in, you sign in first. Check the account shown under Signed in as, then click Approve - or Deny.
  • An API token - for assistants that ask for a token. Create one under API tokens and enter it in the assistant.

To check the connection, ask the assistant to list your CodeLynq projects.

What you allow

AccessWhat the assistant may do
ReadView your projects, tasks and runs
WriteCreate tasks, generate and approve technical designs, start development runs - and the other actions below

With read-only access, any attempt to change something is refused with a message that the connection is read-only.

What the assistant can do

Look things up

  • list your projects, a project's repositories and a repository's branches,
  • check the status of a run,
  • list a project's releases,
  • read a troubleshoot session.

Act (needs write access)

  • create a task - with its type, repositories and branches, including new branches copied from an existing one (see Work on several branches at once),
  • generate a task's technical design, and approve it,
  • start a run,
  • generate or update a project's documentation in a pull request,
  • start a deployment to one of the project's deploy targets,
  • start a troubleshoot session on a task and ask follow-up questions, up to a diagnosis.

Everything goes through the same pipeline as in the app: the same plan limits, spend limits, quality gates and pull requests for review.

How long a connection lasts

A connection made by approving (OAuth) stays valid for 90 days. After that, the assistant asks you to approve again.

Disconnect

Approved assistants appear under Security settings → API tokens as OAuth · {name}, with a Connected app badge. Click Revoke to disconnect one immediately.

Test it on a task from your own backlog

Send us one real task. We run CodeLynq on your repository and you review the pull request. Free, within 5 working days of approving the spec.