@currents/mcp

What it does

The Currents MCP Server connects AI agents directly to your Currents CI dashboard, giving you visibility into test runs, spec files, and individual test results. Agents can query error trends and performance metrics across your test suite, manage quarantine rules and webhooks programmatically, and diagnose flaky tests or failures using actual execution data—no guesswork required.

Who it's for

Backend engineers and QA specialists managing large test suites where flakiness blocks progress. Platform teams running CI/CD systems who want AI agents to understand real test failures instead of static logs. Development teams using Claude or Cursor that need agents to analyze execution data and suggest or implement fixes.

Common use cases

• Query a failing CI run and have Claude diagnose root causes from actual error output and stack traces.
• Identify flaky tests via performance metrics and create quarantine rules programmatically to isolate unstable tests.
• Surface test performance regressions and have agents suggest optimization strategies.
• Generate Jira issues automatically from critical test failures using Currents' Jira integration.
• Reset failed spec files in a run to trigger re-execution, then analyze results after the re-run completes.

Setup pitfalls

• Requires CURRENTS_API_KEY environment variable—obtain it from your Currents dashboard account settings before running.
• Marked high-risk due to filesystem write and network access—run in a sandboxed or trusted development environment only.
• Network and authentication failures fail silently; verify API connectivity and key validity before assuming agent call failures are logic errors.
• Requires an active Currents project with populated test data—agents will have no results without a properly configured project in your account.