What it does
The CircleCI MCP server bridges your CI/CD platform and large language models through Model Context Protocol, enabling natural language queries against your CircleCI projects. You can trigger pipelines, rerun failed jobs, analyze test history, audit configuration, fetch build logs and test results, identify flaky tests, spot resource bottlenecks, and download usage data—all through conversational prompts in Cursor, VS Code, Copilot, Claude, or any MCP client.
Who it's for
DevOps engineers, CI/CD maintainers, and backend developers who need to troubleshoot pipeline failures, optimize compute allocation, or validate CircleCI configuration without context-switching to the web dashboard. Also useful for teams building AI-assisted DevOps tooling.
Common use cases
- Trigger a pipeline or rerun a failed job using a natural language prompt
- Fetch and analyze build failure logs to identify root causes
- Identify flaky tests by querying test execution history across workflow runs
- Validate CircleCI YAML configuration for errors and get remediation suggestions
- Find underused compute resource classes to reduce infrastructure spend
- Generate test cases and evaluate prompt templates for AI applications
Setup pitfalls
- Requires a CircleCI Personal API token (
CIRCLECI_TOKEN) — generate one at app.circleci.com/settings/user/tokens and never commit it to version control. - On-premises CircleCI deployments must set
CIRCLECI_BASE_URL; cloud users can omit it and default to https://circleci.com. - Local NPX installation requires Node.js ≥ v18 and pnpm; Docker installation requires Docker to be running.
- Scanner found one secret in the codebase — review and rotate any exposed tokens before deploying.