Skip to main content

Anthropic Provider

💡TL;DR

Use the Anthropic preset for the native Messages API. A Claude model served by an OpenAI-compatible gateway belongs in that gateway's preset. Confirm current model access before running a task; the historical preset default may need to be changed.

Setup​

Select Anthropic, enter an Anthropic API key and retain the native endpoint unless your deployment explicitly exposes the same protocol. Enter a currently available Claude model ID, run Test connection, and try a disposable note with the intended task.

Endpoint And Authentication​

FieldPreset value
Transportanthropic
API keyRequired
Base URLhttps://api.anthropic.com
Default modelclaude-3-5-sonnet-20240620
Temperature0.5
Connection testchat-only

The runtime uses /v1/messages with the native API-key and API-version headers. Do not put an OpenAI-style chat route into this preset. The model default records a repository setting, not a statement that the upstream provider still serves it.

Model Discovery​

Model discovery uses the native Anthropic model-list route where available. If listing is restricted, enter an accessible model ID and use the chat-based connection test. Verify long or structured tasks separately from the short test.

Troubleshooting​

  • 401 or 403: check key, workspace/project access, billing and model permissions.
  • 404: verify endpoint and current model availability.
  • Gateway model IDs fail: use the matching gateway preset and its expected identifier.
  • Long responses fail: inspect output limits and provider errors before increasing retries.

When To Use​

Choose the native preset when your endpoint serves the Anthropic Messages protocol. Compare model output on your own sources instead of treating provider selection as a universal quality ranking. Task-specific routing lets you use this provider only where it meets your requirements.

Next Steps​