OpenClaw

Use an OpenClaw connection to chat with an agent on your Gateway. The model picker lists agent targets; each agent uses the backend model configured on the Gateway.

Add your Gateway

1. Open Connections, tap +, and choose OpenClaw.
2. Enter a reachable HTTP or HTTPS Gateway address, such as http://YOUR_SERVER:18789/v1. A server root without /v1 also works with the standard route prefix.
3. Enter your Gateway token in API Token. This is the Gateway credential, not your upstream AI provider's API key.
4. Match Use HTTPS to the scheme your Gateway actually serves.
5. Save and fetch the available agents. The app prefers openclaw/default when the Gateway returns it.

Use the Gateway's HTTP API address rather than its WebSocket or dashboard URL. Expand Request addresses to inspect the model and chat URLs. Tokens are stored in Keychain.

Choose an agent

Choose an available agent in Default Model, in a chat's model picker, or in keyboard settings. The default-agent alias follows the Gateway's configured default agent. Selecting an agent in the app does not change that agent's backend model configuration.

If discovery fails

• Check that the device can reach the Gateway address. On iPhone, localhost refers to the phone.
• For authentication errors, check the Gateway token.
• If you receive a missing-endpoint or HTML response, check the address, prefix, and Gateway HTTP Chat Completions configuration.
• An older Gateway may support chat without offering an agent catalog. Existing saved agents can still be used when the catalog is unavailable with a 404 or 405 response; other discovery errors need to be resolved.
• If a named agent disappears, explicitly select an available one. The app preserves your choice instead of silently changing agents.

See Connections for setup basics and Remote Access for reachable server addresses.