Full read-write Claude connector for CircleCI
Let Claude rerun failed workflows and trigger pipelines in CircleCI
Stop tab-switching to babysit builds and rerun after a flake. Claude reads the failing step, proposes the fix, and reruns or triggers the pipeline right where CI already runs, executing the destructive actions only once you approve.
CircleCI is a continuous integration and delivery platform that builds, tests, and ships code from your repositories. Teams want Claude to do more than read pipeline status: to rerun failed jobs, trigger builds, and manage project settings from the conversation.
What is a CircleCI MCP server?
A CircleCI MCP server is a small hosted service that exposes CircleCI's data and actions to Claude over the Model Context Protocol, so Claude can issues, code, and incidents inside your own account. We build it as a custom connector with separate read and write tools, and destructive actions stay gated behind your approval.
What can CircleCI's current connector do today?
CircleCI does publish an official MCP server, but it ships a fixed, standard set of tools: rerun a workflow, trigger a pipeline, read failure logs and status. A custom connector covers the rest of your workflow, like canceling a stuck workflow and managing environment variables, with scoped tokens and human-in-the-loop controls so Claude works inside CI the way your team actually does.
What does full read-write CircleCI access unlock?
Full read-write CircleCI access lets Claude rerun a failed workflow, trigger a pipeline on a branch, cancel a stuck job, and update environment variables once you approve. The build babysitting and rerun-after-flake loops that used to mean tab-switching happen where the pipeline already runs.
- Rerun a failed workflow, optionally from the failed job onward
- Trigger a new pipeline on a specified branch or tag
- Cancel a running or stuck workflow
- Add or update a project environment variable after you approve
- Approve a held job in a workflow that is waiting on a gate
- Surface the failing step's logs and propose the rerun
Can Claude rerun a failed CircleCI workflow?
Yes. Claude can rerun a failed CircleCI workflow with a write tool, either restarting the whole workflow or picking it up from the failed job onward. It surfaces the failing step's logs first and proposes the rerun, so you act on the failure instead of just reading its status.
Can Claude trigger a pipeline on a branch in CircleCI?
Claude can trigger a new pipeline on a specified branch or tag in CircleCI, and approve a held job that is waiting on a gate. Triggering and approval are gated write actions: Claude prepares the request and shows what it will run, then executes only after you confirm.
Can Claude cancel a stuck job or change env vars in CircleCI?
Claude can cancel a running or stuck workflow and add or update a project environment variable in CircleCI, but both are separate, gated actions that run only after you approve. The connector uses a scoped API token, so it never gets more access to CircleCI than your account already has.
How a CircleCI connector works
We build a small MCP server that authenticates to CircleCI with a scoped personal or project API token, never broader than the access the token's owner already has. Read tools (pipeline status, job logs) and write tools (rerun, trigger, cancel, set variables) are separated, and outbound or destructive actions like canceling or changing environment variables require your confirmation. If your CircleCI is behind IP restrictions, the server can run from an allowlisted address.
Frequently asked questions
Can Claude rerun a failed CircleCI workflow, not just read its status?
Can Claude trigger a pipeline or cancel a build without my approval?
Does the connector get more access to CircleCI than I have?
Related connectors
Tell us what you run. We'll build the connector.
Pick your tool, choose what Claude should do, and get an instant scope and price. No call required.