Model providers
Connect Anthropic, OpenAI, Google Gemini, OpenRouter, or managed model access.
Model providers give agents access to the models they need. You can save provider credentials once and reuse them when creating agents.
Supported providers
| Provider | Commonly used with |
|---|---|
| Anthropic | Claude Code |
| OpenAI | Codex |
| Google Gemini | Gemini CLI |
| OpenRouter | Model routing and compatible model access |
Your workspace may also have managed model access. If it does, the creation flow can offer provider options without requiring you to paste a personal API key.
Add a provider key
- Open Settings -> Model providers.
- Select the provider.
- Add a label, such as
personalorteam. - Paste the API key.
- Leave Base URL empty unless you use a compatible custom endpoint.
- Test the connection.
- Save the provider.
When you create an agent, choose the saved provider from the creation flow.
Provider labels
Use labels that explain ownership or budget boundaries:
personalengineering-teamstagingmanaged
Avoid labels that include secret values, customer names, or one-off task names.
Rotating keys
If a key is revoked or replaced:
- Open the provider row in settings.
- Paste the new key.
- Test the connection.
- Save the provider.
- Retry any failed agent action.
Existing agents use the updated provider connection after it is saved.
Troubleshooting
- Test connection fails: confirm the key belongs to the selected provider and has model access.
- Agent creation asks for a provider: add a compatible provider for the selected framework.
- Model list is empty: check account billing, model permissions, and any custom base URL.
- Costs look unexpected: review Settings -> Usage and confirm the agent is using the intended provider.