# 🦞 Claw Setup ### The Ultimate AI Agent Deployment Skill **Setup ANY Claw platform with 25+ AI providers + FREE Qwen OAuth + Full Customization** ---

✨ Autonomously developed by GLM 5 Advanced Coding Model

⚠️ Disclaimer: Test in a test environment prior to using on any live system

---

## Table of Contents 1. [Features Overview](#-features-overview) 2. [Supported Platforms](#-supported-platforms) 3. [FREE Qwen OAuth Import](#-feature-1-free-qwen-oauth-import) 4. [25+ AI Providers](#-feature-2-25-ai-providers) 5. [Customization Options](#-customization-options) 6. [Installation Guides](#-installation-guides) 7. [Configuration Examples](#-configuration-examples) 8. [Usage Examples](#-usage-examples) --- ## 🎯 Features Overview ``` ┌─────────────────────────────────────────────────────────────────┐ │ CLAW SETUP FEATURES │ ├─────────────────────────────────────────────────────────────────┤ │ │ │ ✅ FEATURE 1: FREE Qwen OAuth Cross-Platform Import │ │ • 2,000 requests/day FREE │ │ • Works with ALL Claw platforms │ │ • Qwen3-Coder model (coding-optimized) │ │ • Browser OAuth - no API key needed │ │ │ │ ✅ FEATURE 2: 25+ OpenCode-Compatible AI Providers │ │ • All major AI labs │ │ • Cloud platforms (Azure, AWS, GCP) │ │ • Fast inference (Groq, Cerebras) │ │ • Gateways (OpenRouter: 100+ models) │ │ • Local models (Ollama, LM Studio) │ │ │ │ ✅ FEATURE 3: Full Customization │ │ • Model selection (fetch or custom) │ │ • Security hardening │ │ • Interactive brainstorming │ │ • Multi-provider configuration │ │ │ └─────────────────────────────────────────────────────────────────┘ ``` --- ## 🦀 Supported Platforms | Platform | Language | Memory | Startup | Qwen OAuth | All Providers | Best For | |----------|----------|--------|---------|------------|---------------|----------| | **Qwen Code** | TypeScript | ~200MB | ~5s | ✅ Native | ✅ | FREE coding | | **OpenClaw** | TypeScript | >1GB | ~500s | ✅ Import | ✅ | Full-featured, 1700+ plugins | | **NanoBot** | Python | ~100MB | ~30s | ✅ Import | ✅ | Research, Python devs | | **PicoClaw** | Go | <10MB | ~1s | ✅ Import | ✅ | Embedded, $10 hardware | | **ZeroClaw** | Rust | <5MB | <10ms | ✅ Native | ✅ | Maximum performance | | **NanoClaw** | TypeScript | ~50MB | ~5s | ✅ Import | ✅ | WhatsApp integration | ### Platform Selection Guide ``` ┌─────────────────┐ │ Need AI Agent? │ └────────┬────────┘ │ ▼ ┌───────────────────────┐ │ Want FREE tier? │ └───────────┬───────────┘ ┌─────┴─────┐ YES NO │ │ ▼ ▼ ┌──────────────┐ ┌──────────────────┐ │ ⭐ Qwen Code │ │ Memory limited? │ │ OAuth FREE │ └────────┬─────────┘ │ 2000/day │ ┌─────┴─────┐ └──────────────┘ YES NO │ │ ▼ ▼ ┌──────────┐ ┌──────────┐ │ZeroClaw/ │ │OpenClaw │ │PicoClaw │ │(Full) │ └──────────┘ └──────────┘ ``` --- ## ⭐ FEATURE 1: FREE Qwen OAuth Import ### What You Get | Metric | Value | |--------|-------| | **Requests/day** | 2,000 | | **Requests/minute** | 60 | | **Cost** | **FREE** | | **Model** | Qwen3-Coder (coding-optimized) | | **Auth** | Browser OAuth via qwen.ai | ### Quick Start ```bash # Step 1: Install Qwen Code CLI npm install -g @qwen-code/qwen-code@latest # Step 2: Get FREE OAuth (opens browser for login) qwen --auth-type qwen-oauth -p "test" # Credentials saved to: ~/.qwen/oauth_creds.json # Step 3: Import to ANY platform # ZeroClaw (native provider - auto token refresh) cat > ~/.zeroclaw/config.toml << EOF default_provider = "qwen-oauth" default_model = "qwen3-coder-plus" EOF # Other platforms (OpenAI-compatible) export OPENAI_API_KEY=$(cat ~/.qwen/oauth_creds.json | jq -r '.access_token') export OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" ``` ### Platform-Specific Import #### OpenClaw + FREE Qwen (OpenAI-Compatible) ```bash git clone https://github.com/openclaw/openclaw.git cd openclaw && npm install # Extract token from Qwen OAuth credentials export OPENAI_API_KEY=$(cat ~/.qwen/oauth_creds.json | jq -r '.access_token') export OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" export OPENAI_MODEL="qwen3-coder-plus" npm run start ``` #### NanoBot + FREE Qwen (OpenAI-Compatible) ```bash pip install nanobot-ai # Extract token and configure export OPENAI_API_KEY=$(cat ~/.qwen/oauth_creds.json | jq -r '.access_token') export OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" export OPENAI_MODEL="qwen3-coder-plus" nanobot gateway ``` #### PicoClaw + FREE Qwen (OpenAI-Compatible) ```bash # Extract token and set environment export OPENAI_API_KEY=$(cat ~/.qwen/oauth_creds.json | jq -r '.access_token') export OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" picoclaw gateway ``` #### NanoClaw + FREE Qwen (OpenAI-Compatible) ```bash # Extract token and set environment export OPENAI_API_KEY=$(cat ~/.qwen/oauth_creds.json | jq -r '.access_token') export OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" nanoclaw ``` #### ZeroClaw + FREE Qwen (NATIVE Provider) ```bash # Install ZeroClaw wget https://github.com/zeroclaw-labs/zeroclaw/releases/latest/zeroclaw-linux-amd64 chmod +x zeroclaw-linux-amd64 && sudo mv zeroclaw-linux-amd64 /usr/local/bin/zeroclaw # ZeroClaw has NATIVE qwen-oauth provider support! # First, get OAuth credentials via Qwen Code: qwen && /auth # Select Qwen OAuth → creates ~/.qwen/oauth_creds.json # Configure ZeroClaw to use native qwen-oauth provider cat > ~/.zeroclaw/config.toml << CONFIG default_provider = "qwen-oauth" default_model = "qwen3-coder-plus" default_temperature = 0.7 CONFIG # ZeroClaw reads ~/.qwen/oauth_creds.json directly with auto token refresh! zeroclaw gateway ``` ### Understanding Qwen OAuth Import Methods ``` ┌─────────────────────────────────────────────────────────────────────────────┐ │ QWEN OAUTH IMPORT METHODS │ ├─────────────────────────────────────────────────────────────────────────────┤ │ │ │ METHOD 1: Native Provider (ZeroClaw ONLY) │ │ ───────────────────────────────────────────────────── │ │ • ZeroClaw has built-in "qwen-oauth" provider │ │ • Reads ~/.qwen/oauth_creds.json directly │ │ • Automatic token refresh using refresh_token │ │ • Tracks expiry_date and refreshes when needed │ │ • Configuration: default_provider = "qwen-oauth" │ │ │ │ METHOD 2: OpenAI-Compatible (All Other Platforms) │ │ ───────────────────────────────────────────────────── │ │ • Treats Qwen API as OpenAI-compatible endpoint │ │ • Extract access_token and use as OPENAI_API_KEY │ │ • Set OPENAI_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode/v1 │ │ • Manual re-export needed when token expires │ │ │ │ COMPARISON: │ │ ┌─────────────────┬────────────────┬─────────────────────┐ │ │ │ Feature │ Native │ OpenAI-Compatible │ │ │ ├─────────────────┼────────────────┼─────────────────────┤ │ │ │ Token Refresh │ ✅ Automatic │ ❌ Manual │ │ │ │ Token Expiry │ ✅ Handled │ ⚠️ Re-export needed│ │ │ │ Platforms │ ZeroClaw only │ All others │ │ │ │ Config File │ ~/.qwen/oauth_creds.json │ env vars │ │ │ └─────────────────┴────────────────┴─────────────────────┘ │ │ │ └─────────────────────────────────────────────────────────────────────────────┘ ``` ### OAuth Credentials Structure Qwen Code stores OAuth credentials in `~/.qwen/oauth_creds.json`: ```json { "access_token": "pIFwnvSC3fQPG0i5waDbozvUNEWE4w9x...", "refresh_token": "9Fm_Ob-c8_WAT_3QvgGwVGfgoNfAdP...", "token_type": "Bearer", "resource_url": "portal.qwen.ai", "expiry_date": 1771774796531 } ``` | Field | Purpose | |-------|---------| | `access_token` | Used for API authentication | | `refresh_token` | Used to get new access_token when expired | | `expiry_date` | Unix timestamp when access_token expires | ### API Endpoints | Endpoint | URL | |----------|-----| | **API Base** | `https://dashscope.aliyuncs.com/compatible-mode/v1` | | **Chat Completions** | `https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions` | | **Token Refresh** | `https://chat.qwen.ai/api/v1/oauth2/token` | ### Available Models (FREE Tier) | Model | Best For | |-------|----------| | `qwen3-coder-plus` | Coding (recommended) | | `qwen3-coder-flash` | Fast coding | | `qwen-max` | Complex tasks | --- ## 🤖 FEATURE 2: 25+ AI Providers ### Tier 1: FREE | Provider | Free Tier | Model | Setup | |----------|-----------|-------|-------| | **Qwen OAuth** | ✅ 2,000/day | Qwen3-Coder | `qwen && /auth` | ### Tier 2: Major AI Labs | Provider | SDK Package | Key Models | Features | |----------|-------------|------------|----------| | **Anthropic** | `@ai-sdk/anthropic` | Claude 3.5/4/Opus | Extended thinking, PDF support | | **OpenAI** | `@ai-sdk/openai` | GPT-4o, o1, o3, GPT-5 | Function calling, structured output | | **Google AI** | `@ai-sdk/google` | Gemini 2.5, Gemini 3 Pro | Multimodal, long context | | **xAI** | `@ai-sdk/xai` | Grok models | Real-time data integration | | **Mistral** | `@ai-sdk/mistral` | Mistral Large, Codestral | Code-focused models | ### Tier 3: Cloud Platforms | Provider | SDK Package | Models | Features | |----------|-------------|--------|----------| | **Azure OpenAI** | `@ai-sdk/azure` | GPT-5 Enterprise | Azure integration | | **Google Vertex** | `@ai-sdk/google-vertex` | Claude, Gemini on GCP | Anthropic on Google | | **Amazon Bedrock** | `@ai-sdk/amazon-bedrock` | Nova, Claude, Llama 3 | AWS credentials | ### Tier 4: Aggregators & Gateways | Provider | Models | Features | |----------|--------|----------| | **OpenRouter** | 100+ models | Multi-provider gateway | | **Together AI** | Open source | Fine-tuning, hosting | | **DeepInfra** | Open source | Cost-effective | | **Vercel AI** | Multi-provider | Edge hosting | ### Tier 5: Fast Inference | Provider | Speed | Models | |----------|-------|--------| | **Groq** | Ultra-fast | Llama 3, Mixtral | | **Cerebras** | Fastest | Llama 3 variants | ### Tier 6: Specialized | Provider | Use Case | |----------|----------| | **Perplexity** | Web search integration | | **Cohere** | Enterprise RAG | | **GitLab Duo** | CI/CD AI integration | | **GitHub Copilot** | IDE integration | ### Tier 7: Local/Self-Hosted | Provider | Base URL | Use Case | |----------|----------|----------| | **Ollama** | localhost:11434 | Local model hosting | | **LM Studio** | localhost:1234 | GUI local models | | **vLLM** | localhost:8000 | High-performance serving | --- ## 🎨 Customization Options ### 1. Model Selection **Option A: Fetch from Provider** ```bash # Use included script ./scripts/fetch-models.sh openrouter ./scripts/fetch-models.sh groq ./scripts/fetch-models.sh ollama # Or manually curl -s https://openrouter.ai/api/v1/models \ -H "Authorization: Bearer $KEY" | jq '.data[].id' ``` **Option B: Custom Model Input** ```json { "customModels": { "my-fine-tuned": { "provider": "openai", "modelId": "ft:gpt-4o:org:custom:suffix", "displayName": "My Custom Model" }, "local-llama": { "provider": "ollama", "modelId": "llama3.2:70b", "displayName": "Local Llama 3.2 70B" } } } ``` ### 2. Security Hardening ```bash # Environment variables (never hardcode keys) export ANTHROPIC_API_KEY="your-key" export OPENAI_API_KEY="your-key" # Restricted config permissions chmod 600 ~/.config/claw/config.json chmod 600 ~/.qwen/settings.json # Systemd hardening NoNewPrivileges=true PrivateTmp=true ProtectSystem=strict ``` ### 3. Interactive Brainstorming After installation, customize with brainstorming: | Topic | Questions | |-------|-----------| | **Use Case** | Coding, research, productivity, automation? | | **Model Selection** | Claude, GPT, Gemini, Qwen, local? | | **Integrations** | Telegram, Discord, calendar, storage? | | **Deployment** | Local, VPS, cloud? | | **Agent Personality** | Tone, memory, proactivity? | --- ## 📦 Installation Guides ### Qwen Code (Native FREE OAuth) ```bash npm install -g @qwen-code/qwen-code@latest qwen /auth # Select Qwen OAuth ``` ### OpenClaw ```bash git clone https://github.com/openclaw/openclaw.git cd openclaw && npm install && npm run setup ``` ### NanoBot ```bash pip install nanobot-ai nanobot onboard nanobot gateway ``` ### PicoClaw ```bash wget https://github.com/sipeed/picoclaw/releases/latest/picoclaw-linux-amd64 chmod +x picoclaw-linux-amd64 sudo mv picoclaw-linux-amd64 /usr/local/bin/picoclaw picoclaw gateway ``` ### ZeroClaw ```bash wget https://github.com/zeroclaw-labs/zeroclaw/releases/latest/zeroclaw-linux-amd64 chmod +x zeroclaw-linux-amd64 sudo mv zeroclaw-linux-amd64 /usr/local/bin/zeroclaw zeroclaw gateway ``` --- ## ⚙️ Configuration Examples ### Multi-Provider Setup ```json { "providers": { "qwen": { "type": "oauth", "free": true, "daily_limit": 2000, "model": "qwen3-coder-plus" }, "anthropic": { "apiKey": "${ANTHROPIC_API_KEY}", "baseURL": "https://api.anthropic.com" }, "openai": { "apiKey": "${OPENAI_API_KEY}", "baseURL": "https://api.openai.com/v1" }, "google": { "apiKey": "${GOOGLE_API_KEY}" }, "openrouter": { "apiKey": "${OPENROUTER_API_KEY}", "baseURL": "https://openrouter.ai/api/v1" }, "groq": { "apiKey": "${GROQ_API_KEY}", "baseURL": "https://api.groq.com/openai/v1" }, "ollama": { "baseURL": "http://localhost:11434/v1" } }, "agents": { "free": { "model": "qwen/qwen3-coder-plus" }, "premium": { "model": "anthropic/claude-sonnet-4-5" }, "fast": { "model": "groq/llama-3.3-70b-versatile" }, "local": { "model": "ollama/llama3.2:70b" } } } ``` ### Environment Variables ```bash # ~/.qwen/.env or ~/.config/claw/.env # Qwen OAuth (FREE - from qwen --auth-type qwen-oauth) # Credentials stored in: ~/.qwen/oauth_creds.json export OPENAI_API_KEY=$(cat ~/.qwen/oauth_creds.json | jq -r '.access_token') export OPENAI_BASE_URL="https://dashscope.aliyuncs.com/compatible-mode/v1" export OPENAI_MODEL="qwen3-coder-plus" # Or use paid providers ANTHROPIC_API_KEY=sk-ant-xxx OPENAI_API_KEY=sk-xxx GOOGLE_API_KEY=xxx GROQ_API_KEY=gsk_xxx OPENROUTER_API_KEY=sk-or-xxx MISTRAL_API_KEY=xxx XAI_API_KEY=xxx COHERE_API_KEY=xxx PERPLEXITY_API_KEY=xxx CEREBRAS_API_KEY=xxx TOGETHER_API_KEY=xxx DEEPINFRA_API_KEY=xxx # Cloud providers AZURE_OPENAI_API_KEY=xxx AZURE_OPENAI_ENDPOINT=https://xxx.openai.azure.com/ AWS_ACCESS_KEY_ID=xxx AWS_SECRET_ACCESS_KEY=xxx GOOGLE_CLOUD_PROJECT=my-project GOOGLE_CLOUD_LOCATION=us-central1 ``` --- ## 💬 Usage Examples ### Basic Usage ``` "Setup OpenClaw with FREE Qwen OAuth" "Install NanoBot with all AI providers" "Configure ZeroClaw with Groq for fast inference" ``` ### Advanced Usage ``` "Setup Claw with Anthropic, OpenAI, and FREE Qwen fallback" "Fetch available models from OpenRouter and let me choose" "Configure PicoClaw with my custom fine-tuned model" "Import Qwen OAuth to use with OpenClaw" "Setup Claw platform with security hardening" ``` ### Provider-Specific ``` "Configure Claw with Anthropic Claude 4" "Setup Claw with OpenAI GPT-5" "Use Google Gemini 3 Pro with OpenClaw" "Setup local Ollama models with Claw" "Configure OpenRouter gateway for 100+ models" ``` --- ## 📁 Files in This Skill ``` skills/claw-setup/ ├── SKILL.md # Skill definition (this file's source) ├── README.md # This documentation └── scripts/ ├── import-qwen-oauth.sh # Import FREE Qwen OAuth to any platform └── fetch-models.sh # Fetch models from all providers ``` --- ## 🔧 Troubleshooting ### Qwen OAuth Token Not Found ```bash # Re-authenticate qwen && /auth # Select Qwen OAuth # Check token location ls ~/.qwen/ find ~/.qwen -name "*.json" ``` ### Token Expired ```bash # Tokens auto-refresh in Qwen Code qwen -p "refresh" # Re-export source ~/.qwen/.env ``` ### API Errors ```bash # Verify token is valid QWEN_TOKEN=$(cat ~/.qwen/oauth_creds.json | jq -r '.access_token') curl -X POST "https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions" \ -H "Authorization: Bearer $QWEN_TOKEN" \ -H "Content-Type: application/json" \ -d '{"model": "qwen3-coder-plus", "messages": [{"role": "user", "content": "Hello"}]}' # Check rate limits # FREE tier: 60 req/min, 2000/day ``` ---

Learn more about GLM 5 Advanced Coding Model