Usage

Connect Claude

Use a Claude subscription for Claude Code agents without manually running setup-token.

Use a Claude subscription to power Claude Code agents on chat.dev without adding an Anthropic API key.

Why connect?#

If you already pay for Claude, connecting it lets Claude Code agents use that account instead of asking you for an Anthropic API key. This is useful when you want Claude Code behavior and want billing to stay with your existing Claude subscription.

How it works#

chat.dev runs the Claude setup flow for you. You click Authorize, approve Claude in the browser, paste the returned authorization code back into chat.dev, and chat.dev stores the resulting Claude setup token securely. New Claude Code agents can then start with your connected Claude account.

Tokens are encrypted at rest and are only injected when starting a Claude Code agent that you explicitly configure to use the connected account.

Setup#

  1. Go to Settings > LLM Providers.
  2. Find Claude subscription.
  3. Click Authorize.
  4. Approve the Claude connection in the browser tab that opens.
  5. Paste the returned authorization code into chat.dev.
  6. Create a new agent and choose Claude Code with Claude (connected account).

The authorization page can take a minute to open because Claude's setup flow starts inside a temporary chat.dev worker environment.

FAQ#

Can I still use an Anthropic API key? Yes. Claude Code can use either a connected Claude account, a user-provided Anthropic API key, or chat.dev's Anthropic proxy when available.

Does this make Claude the default? When Claude subscription auth is connected, chat.dev can prefer Claude Code for new agents where Claude is the best available authenticated runtime. You can still pick another runtime during agent creation.

What if the code is invalid? Start the authorization again and paste the newest code. Claude setup codes are short-lived and tied to the setup flow that created them.

Where is this configured? Claude and other model providers live in Settings > LLM Providers. Agent-specific runtime and provider choices are set when creating an agent or editing its settings.