Configure LongCat-2.5 in Claude Code

One-sentence takeaway

LongCat officially provides a Claude Code configuration that uses LongCat-2.5 through an Anthropic-compatible API.

Use cases

• Suitable tasks: Connecting LongCat-2.5 to the Claude Code CLI.

• Unsuitable tasks: Using this configuration to judge the model's performance on specific coding tasks.

• Applicable model version: LongCat-2.5.

• Applicable client, agent, or API: Claude Code CLI and LongCat's Anthropic-compatible API.

• Recommended reasoning level and parameters: The source does not specify a reasoning level; the configuration example sets CLAUDE_CODE_MAX_OUTPUT_TOKENS to 262144.

Ready-to-use content

Create or merge the following configuration in ~/.claude/settings.json, replacing the placeholder with your own LongCat API key:

{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "your_longcat_api_key",
    "ANTHROPIC_BASE_URL": "https://api.longcat.chat/anthropic",
    "ANTHROPIC_MODEL": "LongCat-2.5",
    "ANTHROPIC_SMALL_FAST_MODEL": "LongCat-2.5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "LongCat-2.5",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "LongCat-2.5",
    "CLAUDE_CODE_MAX_OUTPUT_TOKENS": "262144",
    "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": 1
  },
  "permissions": {
    "allow": [],
    "deny": []
  }
}

If you already have a configuration, merge the required fields to avoid overwriting existing settings or permission rules. Do not share a configuration file that contains a real key.

Test or workflow steps

1. Create an account on the LongCat API Platform and manually create a key on the API Keys page.

2. Install Claude Code, configure it using the example above, and enter your key.

3. Run claude in the terminal and enter a simple test message, such as “Hello, please introduce yourself”.

4. Check whether you receive a response from the LongCat model.

Raw evidence and data

• The supported-model table in the official documentation lists LongCat-2.5, with OpenAI / Anthropic as the API formats.

• The official configuration example sets the main model, fast model, and default Sonnet and Opus models all to LongCat-2.5; the API endpoint is https://api.longcat.chat/anthropic.

• The documentation uses receipt of a model response as a basic configuration test, but provides neither the test output nor performance data.

Scope and limitations

This is an official integration configuration, not a complete prompt or an independent performance evaluation. The configuration and supported models may change as the service is updated; check the platform for availability, quotas, and fees at the time of use.

Reproduction notes

You need a LongCat API key and an installed copy of Claude Code. At collection time, the official documentation's configuration and supported-model table were checked; no API request was made with a real key, so an actual response was not verified.