Skip to main content
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.
CometAPI dashboard showing the Add API Key button
CometAPI API key details with base URL
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:
ProviderAPI FormatEndpoint calledBase URL
cometapi-openaiopenai-completionsPOST /v1/chat/completionshttps://api.cometapi.com/v1
cometapi-openai-responsesopenai-responsesPOST /v1/responseshttps://api.cometapi.com/v1
cometapi-claudeanthropic-messagesPOST /messageshttps://api.cometapi.com
cometapi-googlegoogle-generative-aiPOST /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
OpenClaw models status showing CometAPI providers authenticated
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
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.