Your first connection

Connect a model provider

Start Caelis in your workspace and run:

/connect

Choose a supported sign-in flow, configure an API-key provider, or connect a local model provider. Follow the displayed connection instructions.

Then open /model to choose the main model and its supported reasoning effort. Ask a small question about the workspace before starting a larger task.

Connect an ACP agent

Install the external agent separately and make its executable available on the Host’s PATH. In /connect, choose the agent or select Custom for another ACP stdio command.

The selected agent owns its own model connection and execution capabilities. Caelis can only use collaboration and permissions that the agent exposes through ACP.

Add participant roles

Open /team and bind roles to a provider model or ACP agent. Save a configuration you want to reuse. See participants and the external ACP guide.

Connection trouble

Check the executable, arguments, provider configuration, and the selected Host environment. A successful protocol initialization is not the same as an authenticated prompt: verify by completing a small real request.

Accounts and model access

Each provider determines its own login method, available models and charges. The Caelis install does not include credits or a subscription. API credentials belong in the connection flow, not in chat or repository files.

ChatGPT subscription access uses a community-compatible Codex OAuth flow. It is not a documented third-party OpenAI integration; follow the login options shown by your selected connection.

You can use local model providers where supported. Saving sessions locally does not make requests to a remote provider offline.