Coding providers

New to local models?

You can run a model on your own computer with Ollama or LM Studio, then connect it through a coding agent. Follow the first local-model guide for downloads, hardware guidance, and a first-response check. Lyre also has a Set up a local model walkthrough in agent onboarding and Providers settings.

Choose an available provider

A provider supplies the coding assistant; Lyre supplies the workspace and remote workflow. Use a provider exposed by your installed build and authenticate it on the host. Provider accounts, subscriptions, and usage limits are separate from Lyre early access.

The foundation includes provider-specific adapters, including Claude, Codex, and OpenCode. The model list and available controls depend on the adapter, installed tools, credentials, and provider account.

Verify setup before opening a large task

1. Install or configure the selected provider using its own setup instructions.
2. Confirm its authentication works on the host.
3. Select it in Lyre and open the intended project.
4. Ask a small read-only question about that project.
5. Confirm you receive a response before requesting file changes.

If the provider is unavailable, check installation and authentication first. If requests are refused after connecting, check the provider’s quota and account status.

Provider controls are not interchangeable

Model choice, approvals, sandbox settings, and tool permissions follow each provider’s supported options. Do not copy a configuration block from one provider into another. Choose the narrow permissions required by the task.

Adding another provider is a developer integration task. Start with creating an integration; adding a key alone does not create a new adapter.

For multiple keys on one provider and choosing which models are enabled, see Providers, keys and routing.