WeClaw-Proxy

WeChat Open-Platform AI Agent Proxy — Connect any AI Agent to WeChat in one step

中文文档

✨ Features

• 🔌 Multi-Agent Support — OpenAI, DeepSeek, Ollama, Google Gemini, Dify, Coze, Webhook, and Local CLI (Codex / Claude Code / Gemini CLI)
• 📱 Multi-Account WeChat — Bind and manage multiple WeChat accounts simultaneously
• 🧠 Smart Routing — LLM-powered automatic message classification, plus manual /command prefix routing
• 🖥️ Web Admin Panel — Full-featured visual configuration with real-time YAML sync
• 📷 QR Code Login — Scan to bind, with guided nickname setup on success
• 💬 Session Management — Automatic context and conversation history
• 🔍 Gemini Search — Native Google Search integration via Gemini adapter
• 📝 Prompt Variables — Dynamic variables like {cur_date}, {model_id} in system prompts
• 🐳 Cross-Platform — Binary / Docker one-click deploy on Linux / macOS / Windows

📸 Screenshots

Dashboard Overview	Add Agent

QR Code Login	Binding Success

Routing Rules	Multi-Account Management

🚀 Quick Start

Docker (Recommended)

# 1. Create config file
curl -o config.yaml https://raw.githubusercontent.com/amigoer/weclaw-proxy/main/configs/config.example.yaml
# Edit config.yaml with your Agent API Key

# 2. Start
docker run -d \
  --name weclaw-proxy \
  -v ./config.yaml:/data/config.yaml \
  -p 8080:8080 \
  ghcr.io/amigoer/weclaw-proxy:latest

# 3. Open http://localhost:8080 to scan QR code and log in

Binary

Download from Releases:

chmod +x weclaw-proxy-linux-amd64
./weclaw-proxy-linux-amd64 --config config.yaml

Build from Source

git clone https://github.com/amigoer/weclaw-proxy.git
cd weclaw-proxy
make        # Build frontend + Go binary
make dev    # Run in development mode

⚙️ Configuration

server:
  port: 8080

adapters:
  - name: "openai-gpt4"
    type: openai
    api_key: "sk-xxx"
    base_url: "https://api.openai.com/v1"
    model: "gpt-4o"
    system_prompt: "You are a friendly WeChat assistant"

routing:
  default_adapter: "openai-gpt4"
  rules:
    - match:
        prefix: "/claude"
      adapter: "claude"

# Smart Routing (optional)
smart_routing:
  enabled: false
  api_key: "sk-xxx"
  model: "gpt-4o-mini"

💡 All configuration can be edited live in the Web Admin Panel — no need to manually modify YAML files.

Gemini Adapter (with Google Search)

Native Google Gemini API support with optional Google Search tool:

adapters:
  - name: "gemini-search"
    type: gemini
    api_key: "AIza..."
    model: "gemini-2.5-flash-preview-04-17"
    system_prompt: "You are a helpful WeChat assistant"
    extra:
      enable_search: "true"   # Enable Google Search tool

System Prompt Variables

Use dynamic variables in system_prompt — they are automatically replaced at runtime:

Variable	Description	Example
{cur_date}	Current date	2026-03-26
{cur_time}	Current time	22:30:15
{cur_datetime}	Full datetime	2026-03-26 22:30:15
{model_id}	Model ID	gpt-4o
{model_name}	Model name	gpt-4o
{locale}	System locale	zh-CN

system_prompt: "Today is {cur_date}. You are using {model_name}. Reply in {locale}."

### CLI Agent

Directly invoke locally installed AI CLI tools with automatic sub-command inference:

```yaml
adapters:
  - name: "codex"
    type: cli
    base_url: "codex"       # Command path
    extra:
      timeout: "120"        # Timeout in seconds, default 120
      # args: "exec"        # Optional, auto-inferred
      # work_dir: "/project" # Optional, working directory

  - name: "claude"
    type: cli
    base_url: "claude"      # claude -p "message"

  - name: "gemini"
    type: cli
    base_url: "gemini"      # gemini -p "message"

CLI Tool	Command	Auto-Inferred Args
Codex	codex	exec
Claude Code	claude	-p
Gemini CLI	gemini	-p

📦 Supported Platforms

Platform	AMD64	ARM64
Linux	✅	✅
macOS	✅	✅
Windows	✅	✅
Docker	✅	✅

🔗 Links

• Linux.do — Open Source Tech Community

📄 License

MIT