@circleci/mcp-server-circleci

What it does

The CircleCI MCP Server bridges language models and CircleCI, providing 17 tools for natural language interactions with CI/CD workflows. From within an MCP-compatible client like Claude, Cursor, or VS Code, you can trigger and rerun pipelines, retrieve test results and failure logs, download usage data, analyze git diffs against CircleCI configuration rules, identify flaky tests, optimize resource allocation, and manage deployments and rollbacks—all without leaving your development environment.

Who it's for

Developers and DevOps engineers running CircleCI who want CI/CD workflows integrated into their IDE or LLM chat. Teams diagnosing build failures collaboratively, cost-conscious operators analyzing spending and resource use, and engineers automating deployment orchestration benefit most.

Common use cases

• Trigger and rerun pipelines directly from LLM conversations
• Diagnose build failures by retrieving test results and logs
• Identify flaky tests from historical test execution patterns
• Analyze costs and resource optimization via CircleCI usage data
• Validate CircleCI configuration files and troubleshoot setup errors

Setup pitfalls

• Obtain a CircleCI Personal API token from account settings and set CIRCLECI_TOKEN as an environment variable before starting the server
• For CircleCI on-premises deployments, configure CIRCLECI_BASE_URL to match your instance; it defaults to the SaaS endpoint https://circleci.com
• The server reads and writes filesystem data when downloading artifacts and logs; apply appropriate sandbox restrictions in shared or multi-tenant environments
• CI tests are not currently passing; verify the latest npm version functions correctly in a test workflow before using in production pipelines