RepoPulse AI — an MCP-powered autonomous maintainer for open-source repos
RepoPulse AI is an autonomous AI maintainer harness that plugs into any GitHub repository and does the work maintainers drown in: reviewing pull-request diffs, triaging incoming issues, and slicing diffs into security-relevant hunks.
For the Alexa+ track, RepoPulse AI exposes its engine as a Model Context Protocol server over Streamable HTTP (MCP spec 2025-11-25), so any MCP-compatible agent — including Alexa+ skills — can call three tools live:
- review_pull_request_diff — analyzes a raw unified diff and returns structured maintainer remarks, security warnings, and a risk score
- triage_github_issue — classifies an issue, scores severity, suggests labels, and synthesizes a reproduction script
- slice_diff_hunks — decomposes diffs into categorized hunks and flags security-sensitive surfaces
How it works: node dist/cli/index.js serve-mcp-http --port 8787 starts the HTTP transport; every POST to /mcp is dispatched through the existing JSON-RPC handleMessage pipeline (the same battle-tested path as the stdio server). The public endpoint is registered in the Alexa+ skill configuration, and the demo video shows an agent calling all three tools end-to-end.
Built with: TypeScript, Node.js, MCP. License: Apache-2.0. Repo: https://github.com/miriyaladhanwinn/repo-pulse-ai
Log in or sign up for Devpost to join the conversation.