OpenClaw

Connect OpenClaw to CometAPI: install via curl or npm, store your key in ~/.openclaw/.env, configure OpenAI/Anthropic/Google providers, and run 500+ models.

Use CometAPI as the model provider for OpenClaw — 500+ models in one API. This guide covers the current onboarding flow (2026.3.8+).

Prerequisites

• Node.js ≥ 22 — recommended via nvm
• A CometAPI account with an active API key — get yours here

1. Get your CometAPI API key

   Log in to the CometAPI console. Click Add API Key and copy your sk-xxxxx key.

   

   
2. Install or update OpenClaw

   Recommended — use the official installer:

   curl -fsSL https://openclaw.ai/install.sh | bash
   

   Or install via npm:

   npm install -g openclaw@latest
   

   Then run the setup wizard once:

   openclaw onboard --install-daemon
   

   During onboarding, choose Skip in the Model/Auth step if you want to configure CometAPI manually. After onboarding, confirm the gateway is running:

   openclaw gateway status
   openclaw dashboard
   
3. Store your CometAPI key

   Write your key to ~/.openclaw/.env so it is available to the daemon without being hardcoded in config files:

   printf 'COMETAPI_KEY=%s\n' 'your-sk-xxxxx-key' >> ~/.openclaw/.env
   chmod 600 ~/.openclaw/.env
   

   If you already inject COMETAPI_KEY through your shell, service manager, or a secrets workflow, skip this step — the config below uses $\{COMETAPI_KEY\} regardless of how the variable is set.

4. Configure CometAPI providers

   CometAPI supports four API formats. Add one or more provider blocks to ~/.openclaw/openclaw.json:

   Provider	API Format	Endpoint called	Base URL
   cometapi-openai	openai-completions	POST /v1/chat/completions	https://api.cometapi.com/v1
   cometapi-openai-responses	openai-responses	POST /v1/responses	https://api.cometapi.com/v1
   cometapi-claude	anthropic-messages	POST /messages	https://api.cometapi.com
   cometapi-google	google-generative-ai	POST /v1beta/...	https://api.cometapi.com/v1beta

   {
     "agents": {
       "defaults": {
         "model": {
           "primary": "cometapi-claude/your-claude-model-id"
         }
       }
     },
     "models": {
       "mode": "merge",
       "providers": {
         "cometapi-openai": {
           "baseUrl": "https://api.cometapi.com/v1",
           "apiKey": "${COMETAPI_KEY}",
           "api": "openai-completions",
           "models": [{ "id": "your-chat-model-id", "name": "OpenAI-compatible text model" }]
         },
         "cometapi-openai-responses": {
           "baseUrl": "https://api.cometapi.com/v1",
           "apiKey": "${COMETAPI_KEY}",
           "api": "openai-responses",
           "models": [{ "id": "your-responses-model-id", "name": "Responses API model" }]
         },
         "cometapi-claude": {
           "baseUrl": "https://api.cometapi.com",
           "apiKey": "${COMETAPI_KEY}",
           "api": "anthropic-messages",
           "models": [{ "id": "your-claude-model-id", "name": "Claude family model" }]
         },
         "cometapi-google": {
           "baseUrl": "https://api.cometapi.com/v1beta",
           "apiKey": "${COMETAPI_KEY}",
           "api": "google-generative-ai",
           "models": [{ "id": "your-gemini-model-id", "name": "Gemini family model" }]
         }
       }
     }
   }
   

   You can keep only the providers you need, or keep all four to route between formats using provider/model references.

   Replace each your-...-model-id placeholder with a current model ID from the CometAPI Models page that matches the provider block.

   Use cometapi-openai-responses (api: "openai-responses") for models that require the OpenAI Responses API. Use cometapi-openai (api: "openai-completions") for models on the classic Chat Completions path. Both share the same baseUrl (https://api.cometapi.com/v1); the only difference is which downstream endpoint OpenClaw calls.

   The auth-profiles.json file is not required for this setup. Provider keys are resolved from models.providers.*.apiKey via the $\{COMETAPI_KEY\} env var. auth-profiles.json is only needed if you want per-profile credential rotation across multiple keys.

5. Restart and verify

   Restart the gateway after adding .env or changing provider config:

   openclaw gateway restart
   

   Then run diagnostics:

   openclaw models status
   openclaw models list --provider cometapi-claude
   openclaw doctor
   

   A successful setup looks like:

   • openclaw models status resolves the primary model to cometapi-claude/your-claude-model-id with no missing-auth warnings for CometAPI providers
   • openclaw models list --provider cometapi-claude lists your configured models
   • openclaw doctor reports no config or auth errors for CometAPI providers

   
6. Switch models

   # Set the default model in config
   openclaw models set cometapi-claude/your-claude-model-id
   
   # Switch for the current chat session
   /model cometapi-openai/your-chat-model-id
   /model cometapi-openai-responses/your-responses-model-id
   /model cometapi-claude/your-claude-model-id
   /model cometapi-google/your-gemini-model-id
   

Clean reinstall (optional)

If you are fixing a broken older setup, stop the gateway, remove ~/.openclaw, then rerun onboarding before applying the CometAPI config:

openclaw gateway stop
rm -rf ~/.openclaw
openclaw onboard --install-daemon

This removes all local config, sessions, logs, auth state, and workspaces. Then follow Steps 3–5 again.