Skip to main content
A provider is the backend that runs an agent thread: it determines which models are available, how authentication works, and how the model interacts with tools.

​
Supported providers

  • codex (OpenAI / Codex)
  • claude (Anthropic / Claude)
  • opencode (OpenCode)
You can choose a provider per thread, and it’s common to mix providers inside a single team (for example: a Claude lead with Codex implementers).

​
What stays consistent across providers

  • The gg UI: threads, timelines, tool cards, and approvals.
  • Built-in coordination: teams, Team Comms, and delivery tracking.
  • Background work: the Process Manager and Process Monitor.
  • Workspace features: markdown browsing, search, and Git-aware workflows.

​
What changes when you switch providers

  • Model catalog: which models you can pick and how they’re named.
  • Authentication: gg reuses the provider’s CLI login (see Auth pages).
  • Latency / cost profile: different providers optimize differently.
  • “Thinking effort” controls: some providers support an effort/depth setting; others may ignore it.

​
Authentication

gg tries hard to avoid making authentication “a second thing you have to manage.” If you can use a provider successfully from your terminal, gg can usually reuse the same credentials. When that isn’t available (or you just don’t want to think about config paths), some providers also offer in-app sign-in.

​
OpenCode availability (optional)

OpenCode is supported when available on your machine. If you need to explicitly enable/disable it, use the GG_AGENT_OPENCODE_ENABLED environment variable before launching the app.