AI tool setup guides

Before you start

1. 1Create and copy an API Key in Dashboard.
2. 2Use https://api.51cowork.com and select a model through https://api.51cowork.com/v1/models.
3. 3Back up existing client files and merge rather than overwrite providers.
4. 4Test CLI tools in a temporary shell. For desktop tools, save the configuration, select the model, and send a test request.

Claude Code

Claude Code uses the Gateway root and ANTHROPIC_AUTH_TOKEN.

export ANTHROPIC_BASE_URL="https://api.51cowork.com"
export ANTHROPIC_AUTH_TOKEN="<key-from-dashboard>"
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1
export CLAUDE_CODE_ATTRIBUTION_HEADER=0
claude

$env:ANTHROPIC_BASE_URL="https://api.51cowork.com"
$env:ANTHROPIC_AUTH_TOKEN="<key-from-dashboard>"
$env:CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC="1"
$env:CLAUDE_CODE_ATTRIBUTION_HEADER="0"
claude

For user-level persistence, save the following to ~/.claude/settings.json. On Windows use %USERPROFILE%\.claude\settings.json.

{
  "$schema": "https://json.schemastore.org/claude-code-settings.json",
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.51cowork.com",
    "ANTHROPIC_AUTH_TOKEN": "<key-from-dashboard>",
    "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": "1",
    "CLAUDE_CODE_ATTRIBUTION_HEADER": "0"
  }
}

Codex CLI / Extension

Codex uses the Responses wire API.

Merge the provider into ~/.codex/config.toml; Windows uses %USERPROFILE%\.codex\config.toml. Replace the model placeholder with a current gateway model.

model_provider = "51cowork"
model = "<model-from-gateway>"

[model_providers."51cowork"]
name = "51cowork"
base_url = "https://api.51cowork.com"
wire_api = "responses"
requires_openai_auth = true

{
  "OPENAI_API_KEY": "<key-from-dashboard>"
}

• Create the directory if needed.
• Do not remove other providers or settings from an existing config.
• Quit and restart the CLI or extension, then open a new task.

OpenCode

Save this OpenAI-compatible provider to ~/.config/opencode/opencode.json or opencode.jsonc. Use a real model ID from /v1/models.

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "51cowork": {
      "npm": "@ai-sdk/openai",
      "name": "51cowork",
      "options": {
        "baseURL": "https://api.51cowork.com/v1",
        "apiKey": "<key-from-dashboard>"
      }
    }
  }
}

Cursor

1. 1Open model or API settings and enable a custom OpenAI Base URL.
2. 2Enter https://api.51cowork.com/v1 and the API Key from Dashboard.
3. 3Add a model ID returned by /v1/models, select it, and send a test request.

Cherry Studio

1. 1Open Model Services in settings and add a provider named 51cowork with type OpenAI.
2. 2Set API Address to the gateway root https://api.51cowork.com; Cherry Studio appends /v1/chat/completions automatically.
3. 3Enter the API Key from Dashboard, fetch the model list, and add the models you need.
4. 4Enable the provider, select a model in chat, and send a test request.

WorkBuddy

1. 1Open Settings → Model, add a custom model, and choose Custom.
2. 2Enter the full endpoint URL: https://api.51cowork.com/v1/chat/completions.
3. 3Enter the API Key from Dashboard and an exact model ID returned by /v1/models.
4. 4Enable Custom Protocol in advanced settings to use the full URL without appending a path.
5. 5Save, select the custom model in chat, send a test request, and check Dashboard usage.

Update a Key

1. 1Create or select a replacement Key in Dashboard.
2. 2Update every environment or user config with the new Key.
3. 3Fully restart clients that cache credentials.
4. 4Send a minimal request and verify it in Usage.
5. 5Delete an unused Key only after the migration is complete.

Configuration acceptance checklist

• Gateway root is https://api.51cowork.com; OpenAI SDK Base URL is https://api.51cowork.com/v1.
• The Key has no extra spaces, quotes, or line breaks.
• The model ID comes from /v1/models.
• Client protocol matches the selected endpoint.
• The successful request appears in Dashboard Usage.