Wechat Devtools MCP

wechat-devtools-mcp

Skill by ara.so — Devtools Skills collection.

Overview

wechat-devtools-mcp wraps the WeChat Developer Tools CLI as an MCP (Model Context Protocol) server, enabling AI coding agents to control the WeChat IDE programmatically. It provides 7 aggregated tools covering the full mini program lifecycle: IDE management, build/deploy, automated testing, debugging, screenshots, and file operations.

Architecture: Thin MCP (7 tools) + Fat Skill (SOPs, parameter references, guardrails). The Skill is required — without it, AI agents cannot execute standard workflows correctly.

Platform: Cross-platform (Windows/macOS). Published to official MCP Registry.

Installation

1. Install MCP Server

# Install uv if not present
pip install uv

# Install wechat-devtools-mcp globally
uv tool install wechat-devtools-mcp --force

Upgrade:

# Kill running instances first
taskkill /F /IM "wechat-devtools-mcp*" 2>/dev/null
uv tool upgrade wechat-devtools-mcp

2. Enable WeChat IDE Service Port

Critical: Open WeChat Developer Tools → Settings → Security → Service Port → Enable.

Verify with wechat_ide(action='status') — connection failure means the port is disabled.

3. Configure MCP Client

Claude Desktop / Antigravity (claude_desktop_config.json):

{
  "mcpServers": {
    "wechat-devtools": {
      "command": "uvx",
      "args": ["wechat-devtools-mcp"],
      "env": {
        "WECHAT_DEVTOOLS_CLI": "C:\\Program Files (x86)\\Tencent\\微信web开发者工具\\cli.bat",
        "WECHAT_PROJECT_PATH": "D:\\Projects\\my-miniapp"
      }
    }
  }
}

macOS (Claude Code .mcp.json in project root):

{
  "mcpServers": {
    "wechat-devtools": {
      "command": "/opt/homebrew/bin/uvx",
      "args": ["wechat-devtools-mcp"],
      "env": {
        "PATH": "/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin",
        "WECHAT_DEVTOOLS_CLI": "/Applications/wechatwebdevtools.app/Contents/MacOS/cli",
        "WECHAT_PROJECT_PATH": "/Users/you/Projects/my-miniapp",
        "NODE_PATH": "/opt/homebrew/bin/node"
      }
    }
  }
}

Environment Variables:

• WECHAT_DEVTOOLS_CLI: Absolute path to CLI (cli.bat on Windows, cli on macOS)
• WECHAT_PROJECT_PATH: Absolute path to mini program project root
• Windows: Escape backslashes (\\), macOS: Use forward slashes

4. Install Skill (Required)

Claude Code:

npx -y skills add WaterTian/wechat-devtools-mcp/.agents/skills/wechat-devtools

Other clients (place in .agents/skills/):

git clone --depth 1 https://github.com/WaterTian/wechat-devtools-mcp.git .wdm-tmp
mkdir -p .agents/skills
cp -r .wdm-tmp/.agents/skills/wechat-devtools .agents/skills/
rm -rf .wdm-tmp

Result:

project/
└── .agents/skills/
    └── wechat-devtools/
        ├── SKILL.md                # Main SOPs + capability map
        └── references/
            └── tool_reference.md   # Full API parameter docs

Core Tools

1. wechat_ide — IDE Lifecycle

Actions: open, login, is_login, close, quit, status

# Open IDE and load project
wechat_ide(action='open')

# Check login status
wechat_ide(action='is_login')

# Get IDE connection status + MCP version
wechat_ide(action='status')

# Close project (keeps IDE running)
wechat_ide(action='close')

# Quit IDE entirely
wechat_ide(action='quit')

Key patterns:

• Always call status first to verify IDE connectivity
• open is idempotent — won't fail if already open
• Use close between test sessions, quit only when necessary

2. wechat_build — Build & Deploy

Actions: compile, preview, upload, build_npm, cache_clean

# Full compilation
wechat_build(action='compile')

# Generate preview QR code
result = wechat_build(
    action='preview',
    extra_args={
        'qr_format': 'terminal',  # or 'image', 'base64'
        'qr_output': '/tmp/qr.png',
        'compile_condition': '{"pathName":"pages/index/index"}'
    }
)

# Upload to WeChat backend (production)
wechat_build(
    action='upload',
    extra_args={
        'version': '1.0.0',
        'desc': 'Initial release'
    }
)

# Build npm dependencies
wechat_build(action='build_npm')

# Clean cache before rebuild
wechat_build(action='cache_clean')

QR formats:

• terminal: ASCII art in console
• image: Save to qr_output path
• base64: Data URI string

3. wechat_automator — Automated Testing

Actions: start, tap, input, element_info, set_data, call_method, call_wx, mock_wx, evaluate, page_stack, page_data, system_info, storage

# Start automation session
wechat_automator(action='start')

# Tap element by selector
wechat_automator(
    action='tap',
    extra_args={
        'selector': '.login-btn',
        'wait_for': 2000  # Wait 2s after tap
    }
)

# Input text
wechat_automator(
    action='input',
    extra_args={
        'selector': 'input.username',
        'text': 'testuser'
    }
)

# Get element properties
wechat_automator(
    action='element_info',
    extra_args={'selector': '.status-text'}
)

# Mock WeChat API
wechat_automator(
    action='mock_wx',
    extra_args={
        'api': 'request',
        'result': 'success',
        'data': '{"code": 200, "data": {"user": "mock"}}'
    }
)

# Get page data
wechat_automator(action='page_data')

# Execute JavaScript
wechat_automator(
    action='evaluate',
    extra_args={'code': 'getCurrentPages()[0].data.userInfo'}
)

Selector syntax:

• .class-name — Class selector
• #id — ID selector
• view.item[data-id="123"] — Attribute selector
• Use >>> for shadow DOM: custom-component >>> .inner-element

4. wechat_inspector — Runtime Logs

Actions: console, cdp

# Capture console logs (10 seconds)
logs = wechat_inspector(
    action='console',
    extra_args={'duration': 10}
)

# Capture Chrome DevTools Protocol events
cdp_logs = wechat_inspector(
    action='cdp',
    extra_args={
        'duration': 5,
        'events': ['Network.requestWillBeSent', 'Runtime.consoleAPICalled']
    }
)

Use cases:

• Detect runtime errors before they crash
• Monitor network requests during user flow
• Capture console.log/warn/error for debugging

5. wechat_screenshot — Visual Testing

# Screenshot current page
wechat_screenshot(extra_args={'full_page': True})

# Screenshot specific element
wechat_screenshot(extra_args={
    'selector': '.product-list',
    'full_page': False
})

Returns base64-encoded PNG. For long pages, automatically stitches scrolling captures.

6. wechat_navigate — Navigation + Log Capture

# Navigate to page and capture CDP logs
wechat_navigate(extra_args={
    'page': 'pages/detail/detail',
    'query': 'id=123',
    'log_duration': 3
})

Combines wx.navigateTo() + wechat_inspector(cdp) in one call.

7. wechat_file — Project Introspection

Actions: project_info, list_pages, read_page, read_file

# Get project.config.json
wechat_file(action='project_info')

# List all pages in app.json
wechat_file(action='list_pages')

# Read page source (WXML + WXSS + JS + JSON)
wechat_file(
    action='read_page',
    extra_args={'page_path': 'pages/index/index'}
)

# Read arbitrary file
wechat_file(
    action='read_file',
    extra_args={'file_path': 'utils/request.js'}
)

Standard Operating Procedures (SOPs)

SOP A: Initial Setup Verification

1. wechat_ide(status) — Verify IDE connectivity
2. wechat_file(project_info) — Confirm project loaded
3. wechat_ide(is_login) — Check login status
4. If not logged in: wechat_ide(login) and wait for user scan

SOP B: UI Debugging Workflow

1. wechat_automator(start) — Begin session
2. wechat_navigate(page='target/page', log_duration=3)
3. wechat_screenshot(full_page=True) — Capture UI state
4. wechat_inspector(console, duration=5) — Check for errors
5. wechat_automator(page_data) — Inspect data bindings

SOP C: Error Investigation

1. wechat_inspector(console, duration=10) — Capture logs
2. wechat_automator(page_stack) — Check navigation state
3. wechat_file(read_page, page_path=<current>) — Review source
4. wechat_automator(evaluate, code='getApp().globalData') — Check global state

SOP D: Full Page Health Check

pages = wechat_file(action='list_pages')
for page in pages['pages']:
    wechat_navigate(extra_args={'page': page, 'log_duration': 2})
    logs = wechat_inspector(action='console', extra_args={'duration': 2})
    if 'error' in logs.lower():
        print(f"❌ {page}: {logs}")
    else:
        print(f"✅ {page}: OK")

SOP E: Mock-Based Integration Test

1. wechat_automator(start)
2. wechat_automator(mock_wx, api='request', result='success', data=<mock_json>)
3. wechat_automator(tap, selector='.trigger-request-btn')
4. wechat_inspector(console, duration=3) — Verify mock response handling
5. wechat_automator(page_data) — Check UI updated correctly

SOP F: Deployment Workflow

1. wechat_build(cache_clean) — Fresh build
2. wechat_build(compile) — Check for errors
3. If errors: Stop and report
4. wechat_build(preview) — Generate test QR
5. User scans and validates
6. wechat_build(upload, version=<semver>, desc=<changelog>)

SOP G: Page Parameter Discovery

# Find page query params from source
source = wechat_file(action='read_page', extra_args={'page_path': 'pages/detail/detail'})
# Parse onLoad(options) in JS file
# Extract options keys: id, type, etc.

Configuration Patterns

Multi-Project Setup

For multiple mini programs, create project-specific .mcp.json:

{
  "mcpServers": {
    "wechat-shop": {
      "command": "uvx",
      "args": ["wechat-devtools-mcp"],
      "env": {
        "WECHAT_DEVTOOLS_CLI": "C:\\...\\cli.bat",
        "WECHAT_PROJECT_PATH": "D:\\Projects\\shop-miniapp"
      }
    },
    "wechat-admin": {
      "command": "uvx",
      "args": ["wechat-devtools-mcp"],
      "env": {
        "WECHAT_DEVTOOLS_CLI": "C:\\...\\cli.bat",
        "WECHAT_PROJECT_PATH": "D:\\Projects\\admin-miniapp"
      }
    }
  }
}

Switch projects by calling the corresponding server's tools.

Custom Compile Conditions

# Target specific page on preview
wechat_build(
    action='preview',
    extra_args={
        'compile_condition': json.dumps({
            'pathName': 'pages/cart/cart',
            'query': 'from=share&id=456',
            'scene': 1007  # WeChat group share
        })
    }
)

Troubleshooting

"Connection refused" on all commands

Cause: Service port not enabled in IDE.

Fix: Settings → Security → Service Port → Enable. Restart IDE.

wechat_automator commands timeout

Cause: Page not fully loaded or selector invalid.

Fix:

1. Add wait_for delay: {'selector': '.btn', 'wait_for': 2000}
2. Verify selector with wechat_automator(element_info)
3. Check page stack: wechat_automator(page_stack)

Preview QR not generating

Cause: Compilation errors or missing qr_format.

Fix:

1. Run wechat_build(compile) first
2. Specify qr_format: 'terminal', 'image', or 'base64'
3. For image, provide qr_output absolute path

Screenshots are blank

Cause: Page render incomplete or IDE minimized.

Fix:

1. Add 2-3s delay before screenshot: wechat_navigate() then wait
2. Ensure IDE window is visible (not minimized)
3. Check if selector exists: wechat_automator(element_info)

Mock not working

Cause: API name typo or wrong mock timing.

Fix:

1. Use exact API name: request, getStorage, showToast (case-sensitive)
2. Call mock_wx before triggering the API
3. Verify with wechat_inspector(console) to see actual API calls

File operations return "not found"

Cause: Incorrect path format or file outside project.

Fix:

• Use forward slashes: pages/index/index
• No leading slash: utils/api.js not /utils/api.js
• Path relative to WECHAT_PROJECT_PATH root

Best Practices

1. Always start with status: Verify IDE connectivity before workflows
2. Log everything: Capture console/CDP logs before and after critical operations
3. Use close not quit: Preserve IDE state between test runs
4. Mock early: Set up mock_wx before navigating to pages
5. Screenshot + logs: Combine visual + text evidence for bug reports
6. Clean cache on errors: Run cache_clean + compile if builds behave oddly
7. Version everything: Use semantic versioning in upload action
8. Test preview first: Never upload without validating via preview QR
9. Read source before mocking: Use read_page to understand data flow
10. Automate health checks: Run SOP D daily to catch regressions early

Integration with Other MCPs

• CloudBase MCP: Use for cloud functions/database (replaces deprecated wechat_cloud)
• Chrome DevTools MCP: Cross-reference CDP logs for H5 pages
• File system MCP: Batch edit mini program source files

Example: Complete Bug Fix Workflow

# 1. Verify setup
status = wechat_ide(action='status')
assert status['connected']

# 2. Navigate to buggy page
wechat_navigate(extra_args={
    'page': 'pages/order/order',
    'query': 'id=789',
    'log_duration': 5
})

# 3. Capture initial state
screenshot_before = wechat_screenshot(extra_args={'full_page': True})
logs_before = wechat_inspector(action='console', extra_args={'duration': 3})

# 4. Read source to identify issue
source = wechat_file(action='read_page', extra_args={'page_path': 'pages/order/order'})
# (Agent analyzes source, suggests fix)

# 5. Apply fix (via separate file edit MCP)
# ...

# 6. Verify fix
wechat_build(action='compile')
wechat_navigate(extra_args={'page': 'pages/order/order', 'query': 'id=789'})
screenshot_after = wechat_screenshot(extra_args={'full_page': True})
logs_after = wechat_inspector(action='console', extra_args={'duration': 3})

# 7. Compare before/after
assert 'error' not in logs_after.lower()

# 8. Deploy
wechat_build(action='preview', extra_args={'qr_format': 'terminal'})
# User validates, then:
wechat_build(action='upload', extra_args={'version': '1.0.1', 'desc': 'Fix order page crash'})

Version Compatibility

• MCP Server: v0.9.7+
• WeChat DevTools: 1.06.2302270+ (stable channel)
• Python: 3.8+ (automatically managed by uv)
• Node.js: 14+ (for npx skills command)

Check current MCP version: wechat_ide(action='status')['mcp_version']

Additional Resources

• Full API Reference
• Skill Source
• WeChat CLI Official Docs
• MCP Registry Entry