Skip to main content
Watch: I Built an AI App Without Leaving Databricks

Setup guide - From DeepSeek and Goose docs

Use DeepSeek in Goose

Goose has no built-in DeepSeek provider, but it documents custom providers that use OpenAI-compatible APIs (Desktop, goose configure, or a JSON file in custom_providers). You point one at DeepSeek's documented chat completions endpoint with model deepseek-flash.

Last verified 4 steps4 official sources

Steps

  1. Step 1.

    goose Desktop: Settings > Models > Configure providers > Add Custom Provider. Set Provider Type to OpenAI Compatible, API URL to DeepSeek's endpoint, API Key to your DeepSeek key, and Available Models to deepseek-flash, deepseek-v4-pro. Then click Create Provider.

  2. Step 2.

    goose CLI: run goose configure, choose Custom Providers > Add A Custom Provider, and enter the same details (API Type: OpenAI Compatible).

    Terminal
    goose configure
  3. Step 3.

    Config file option: create ~/.config/goose/custom_providers/custom_deepseek.json (Windows: %APPDATA%\Block\goose\config\custom_providers\). This follows Goose's documented schema, filled with DeepSeek values.

    JSON
    {
  "name": "custom_deepseek",
  "engine": "openai",
  "display_name": "DeepSeek",
  "description": "DeepSeek API",
  "api_key_env": "CUSTOM_DEEPSEEK_API_KEY",
  "base_url": "https://api.deepseek.com/chat/completions",
  "models": [
    { "name": "deepseek-flash", "context_limit": 1048576 },
    { "name": "deepseek-v4-pro", "context_limit": 1048576 }
  ],
  "supports_streaming": true,
  "requires_auth": true
}
  4. Step 4.

    Set the key named in api_key_env and start a session with the provider.

    Terminal
    export CUSTOM_DEEPSEEK_API_KEY="your-api-key"
goose session start --provider custom_deepseek

Caveats

  • This combines Goose's documented custom-provider mechanism and JSON schema with DeepSeek's documented endpoint (curl https://api.deepseek.com/chat/completions) and model ids. The provider name custom_deepseek and the env var name are our own choices.
  • context_limit 1048576 matches DeepSeek's documented 1M context (the same value DeepSeek uses in its Codex models.json).
  • Alternative documented path: Goose's built-in OpenAI provider with OPENAI_HOST=https://api.deepseek.com and OPENAI_BASE_PATH=chat/completions. Goose's docs say some endpoints are mounted at chat/completions with no v1.
  • DeepSeek thinking mode is on by default. Goose relies on tool calling, and requests that include tools must pass reasoning_content back or the API returns a 400 error. Goose's docs do not say whether custom OpenAI-compatible providers do this.

More setup guides

Related reading

FAQ

How do I use DeepSeek in Goose?

Goose has no built-in DeepSeek provider, but it documents custom providers that use OpenAI-compatible APIs (Desktop, goose configure, or a JSON file in custom_providers). You point one at DeepSeek's documented chat completions endpoint with model deepseek-flash.

Which DeepSeek model ids work in Goose?

The official docs read for this guide list deepseek-flash and deepseek-v4-pro. Requests go to https://api.deepseek.com/chat/completions.

Is this an official DeepSeek and Goose setup?

The steps draw on both DeepSeek's and Goose's docs. It was last checked against 4 official pages on October 17, 2026.

Sources

Last verified against 4 official pages.

Newsletter

New models, set up the day they ship

One email per week. New model releases, the coding tools that support them and how to wire them up.