chaixueyuan/weapp-agent-mcp

This page was translated automatically. Technical details may differ from the original. View the original

22 stars 0 forks Quality 75 Trend 14MIT

Agent-first MCP server for WeChat Mini Program debugging & automation. The WeChat Mini Program MCP service for agents allows Claude Code / Cursor to directly operate WeChat developer tools.

Overview

WeChat Mini Program MCP Service for Agents — Allows agents like Claude Code / Cursor / Codex to directly drive WeChat developer tools in conversations for debugging, UI operations, page inspection, and lightweight testing. 📖 Changelog · 🐛 Issues · 💬 Discussions weapp-agent-mcp is an agent-oriented WeChat Mini Program MCP service, encapsulating WeChat developer tool automation capabilities based on miniprogram-automator, used for page debugging, element manipulation, lightweight regression testing, and recovery-friendly troubleshooting. It is suitable for: - Agent main development / main debugging - Serial screenshots and page inspection - Lightweight scenarios / reports - Automated verification based on stable selector If you're already using agents like Claude Code / Cursor / Codex to write small programs, traditional miniprogram-automator script calls aren't smooth enough — you have to have agents write node scripts, run, read output, and then adjust. This project directly exposes automation capabilities as agent tool calls, allowing agents to can: - 🤖 — Directly calls mp_screenshot / page_snapshot / element_tap, no need to transfer scripts - 🔁 — Built-in mp_healthCheck / mp_recoverConnection, self-heals after disconnection, longer sessions more stable - 📋 — mp_runScenario + mp_generateScenarioReport generates markdown reports, suitable for PR/oncall - 🎯 — Designed to encourage qa-* named anchor points to reduce script fragmentation after UI redesign - WeChat Developer Tools installed, supporting command-line access (CLI / cli.bat) - Node.js 18+ and npm are installed locally - Mini Program projects that can be opened in Developer Tools - Automated testing and service ports have been enabled in WeChat Developer Tools You don't need to pull the repository to your local site first, nor do you need to manually build a dist. As long as your machine has Node.js and npx locally, you can configure it directly in the MCP client:

Overview (machine translated)

Agent-first MCP server for WeChat Mini Program debugging, automation, and lightweight regression testing.

Agent-oriented WeChat Mini Program MCP service — Allows agents like Claude Code / Cursor / Codex to directly drive WeChat developer tools in conversations for debugging, UI operations, page inspection, and lightweight testing.

📖 Changelog · 🐛 Issues · 💬 Discussions

weapp-agent-mcp is an agent-oriented WeChat Mini Program MCP service, encapsulating WeChat Developer Tool automation capabilities based on miniprogram-automator, used for page debugging, element manipulation, lightweight regression testing, and recovery-friendly troubleshooting.

It is suitable for:

agent: Main development / main debugging
Serial screenshots and page patrol
Lightweight scenario / report
Automated verification based on stable selector

It should not currently be promoted as:

Concurrent screenshot channel
Ultra-long high-voltage full-link return-back engine

For a more complete document, see docs/README.md.

Why Choose weapp-agent-mcp

If you’re already using agents like Claude Code / Cursor / Codex to write small programs, traditional miniprogram-automator script-based calls aren’t smooth enough — you have to have agents write node scripts, run, read output, and then adjust. This project exposes automation capabilities directly as agent tool calls, allowing agents to can:

🤖 Zero Script Debugging — Directly call mp_screenshot / page_snapshot / element_tap, no need to relay scripts
🔁 Restore friendliness — Built-in mp_healthCheck / mp_recoverConnection], can self-heal after disconnection, making long conversations more stable
📋 Lightweight Regression — mp_runScenario + mp_generateScenarioReport generates markdown reports, suitable for PR/oncall
🎯 Stable selector priority — The design encourages qa-* naming anchors to reduce script fragmentation after UI redesigns

Applicable / Not Applicable Comparison

Scenario	Is it applicable
Agent Active Driving Debugging / Inspection	✅ Main Recommended Scenarios
Writing Traditional Scripted E2E Testing	⚪ Directly using the native `miniprogram-automator` SDK is lighter
Concurrent screenshot / stress test	❌ Not applicable — screenshot channel designed as serial

This page was translated automatically. Technical details may differ from the original.

Install

npx -y @chaixueyuan/weapp-agent-mcp

New to MCP? Start with the mcpskill.org guide →

Configuration

{
  "mcpServers": {
    "weapp-agent-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@chaixueyuan/weapp-agent-mcp"
      ],
      "env": {
        "WEAPP_WS_ENDPOINT": "ws://localhost:9420"
      }
    }
  }
}