Use BuildPulse in Claude, ChatGPT & Cursor
BuildPulse ships a Model Context Protocol (MCP) server, so your AI assistant can query your flaky tests, test results, and CI analytics directly, with no copy-pasting logs. Ask “which tests are flakiest this week?” and get a real answer from your data.
What you can ask
- “Which tests are flakiest in this repo right now, and how often do they fail?”
- “Show me the failing tests from my last CI run and the likely cause.”
- “What’s our test pass rate trend over the last 30 days?”
- “Quarantine the top flaky test so it stops blocking merges.”
- “Did the suite get faster after we moved this job to BuildPulse runners? Compare per-test timings before and after.”
The server exposes what BuildPulse records about your CI: recent CI runs and their per-test results and timings (on BuildPulse runners or any other runner), flaky-test history, repository flakiness, and coverage. It does not expose runner minutes or billing; that lives on your usage page.
Hosted setup (recommended)
For clients that support remote MCP servers, point them at the hosted endpoint and sign in with BuildPulse (OAuth):
{
"mcpServers": {
"buildpulse": {
"url": "https://mcp.buildpulse.io"
}
}
}Endpoint: https://mcp.buildpulse.io · Full tool reference in the MCP server docs.
Local setup (stdio)
For Claude Desktop, Cursor, and other stdio clients, add the npm package to your MCP config:
{
"mcpServers": {
"buildpulse": {
"command": "npx",
"args": ["-y", "@buildpulse/mcp"]
}
}
}Package: @buildpulse/mcp · Also listed on the official MCP Registry as io.github.BuildPulseLLC/buildpulse-mcp.
You need a BuildPulse account to connect
The MCP server queries your own test data: flaky tests, recent failures, coverage. Create a free account, connect a repo, and your assistant can start asking questions about your CI.
Free to start · No credit card required