node, and a Limitless account with owner or admin access to a workspace. Local integration testing used OpenCode v1.18.21; the integration contract was also checked against v1.19.0.
Connect
Run this in your terminal:- Your browser opens Limitless. Sign in if needed. If it does not open, follow the link printed in your terminal.
- Check that the browser shows the same connection code as your terminal. Only approve a connection you started yourself.
- Select the workspace and active project that should pay for requests, then choose Approve connection.
- Return to the terminal and wait for OpenCode to confirm the connection. Start a new OpenCode session to load the configuration.
/models to use DeepSeek. DeepSeek Flash is the default for new connections; existing user or project OpenCode settings take precedence.
Newly available models appear when you start a new OpenCode session. Existing connections do not need to reconnect or edit their configuration. Disabled models and models awaiting required pricing are excluded automatically. Coding-tool output budgets are capped at 32,768 tokens or the model’s smaller runtime limit.
Your key and usage
The connection creates a dedicated key named OpenCode, valid for 90 days. It has the same inference permissions as a normal project API key; it is not restricted to GLM. Usage belongs to your selected project and uses the workspace’s available usage, spending caps and request limits. Choose a monthly plan or buy prepaid API usage in Billing before making requests. If you reach your monthly limit, wait for the next reset, upgrade, or buy extra usage. Connecting OpenCode does not start a subscription or collect payment. Manage or revoke the key in API keys. After revoking or expiring a key, run the connection command again to create a replacement.If connection or inference fails
- Expired code: run the command again and use the new browser link. A request lasts ten minutes.
- Interrupted connection: check the terminal’s connection message; a zero exit code alone does not establish success. If the terminal stopped after your approval, inspect API keys and revoke any new unused OpenCode key before retrying. A key may have been issued before OpenCode saved it.
- Insufficient permissions: choose a workspace you own or administer and an active project.
- Project policies: GLM cannot evaluate published project policies. Choose a project without applicable policies.
- Model unavailable: a successful connection confirms your credentials. Model availability is separate; if the model is temporarily unavailable, try again later. Requests do not automatically fall back to another provider.
- Different model or endpoint: local OpenCode settings take precedence over the supplied defaults. Review user and project configuration before retrying.