Weapp Agent Mcp

Bridges Claude and other AI agents directly to WeChat DevTools via WebSocket, wrapping miniprogram-automator into MCP tool calls. Instead of asking the agent to write Node scripts for debugging, you get direct access to mp_screenshot, page_snapshot, element_tap, and assertion tools like page_expectVisible. Built for serial UI checks and lightweight regression flows, not parallel screenshot stress tests. Includes mp_healthCheck and mp_recoverConnection to self-heal through long debugging sessions. Designed around stable qa-* selectors to survive UI refactors. Best when agents are already driving your mini program development and you want them debugging in the same conversational flow. Can run pre-built scenario suites and generate markdown reports for PRs.