Configure LongCat-2.5 in OpenClaw

One-sentence takeaway

LongCat's official documentation lists OpenClaw as supporting LongCat-2.5 and provides configuration and verification steps for the OpenAI-compatible API.

Use cases

• Suitable tasks: Connecting LongCat-2.5 in OpenClaw as the default model.

• Unsuitable tasks: Assessing model performance or configuration compatibility with other clients based on this guide.

• Applicable model version: LongCat-2.5.

• Applicable client, agent, or API: OpenClaw, using the LongCat OpenAI-compatible API.

• Recommended reasoning tier and parameters: The example sets reasoning: false, contextWindow: 1048576, and maxTokens: 262144; the documentation does not recommend other reasoning tiers.

Ready-to-use content

Merge the following fields into ~/.openclaw/openclaw.json, keeping your existing configuration. Replace the placeholder with your own LongCat API Key. Do not commit a real key to the repository.

{
  "models": {
    "mode": "merge",
    "providers": {
      "longCat": {
        "baseUrl": "https://api.longcat.chat/openai/v1",
        "apiKey": "YOUR_API_KEY",
        "api": "openai-completions",
        "authHeader": true,
        "models": [
          {
            "id": "LongCat-2.5",
            "name": "LongCat-2.5",
            "reasoning": false,
            "input": ["text"],
            "contextWindow": 1048576,
            "maxTokens": 262144,
            "compat": { "maxTokensField": "max_tokens" }
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": { "primary": "longCat/LongCat-2.5" }
    }
  }
}

You can also add a provider in the Web control panel under Config → Models → Providers: select openai-completions for API, enter https://api.longcat.chat/openai/v1 as the Base URL, set the model ID and name to LongCat-2.5, and enter your API Key.

Test or workflow steps

1. Install OpenClaw, prepare a LongCat API Key, and make sure your account has sufficient balance.

2. Run openclaw onboard --install-daemon to initialize. The official guide recommends selecting Skip for now for the initial model provider.

3. Merge the configuration above and save it. The official guide says it takes effect immediately after saving.

4. Run openclaw tui and use /status to check the Gateway; or run openclaw dashboard to open the Web UI, send a test message on the Chat page, and confirm that the model responds.

Original evidence and data

• The official supported-model table lists LongCat-2.5 as supported, with OpenAI / Anthropic API formats; this article's configuration uses the OpenAI-compatible interface given in the documentation.

• The documentation lists ~/.openclaw/openclaw.json as the configuration file location and http://127.0.0.1:18789 as the default Web address.

• The documentation's separate default-model configuration uses longCat/LongCat-2.5, while its full configuration example still sets primary to the older model longCat/LongCat-Flash-Thinking-2601. This article uses the former, which matches the target model.

Scope and limitations

This is an official configuration guide, not a model performance test; it has not been independently verified. Parameters and the OpenClaw interface may change between versions, so refer to the original and your installed version. reasoning: false is an example configuration and does not imply that the model lacks reasoning capabilities.

Reproduction notes

Open the original and review Supported Models, LongCat Model Configuration, and Getting Started in order. Use your own key when configuring, then send a test message to confirm that the call succeeds.