Skip to main content
Applies to:
  • Plan -
  • Deployment -

​
Summary

Loop shows models that are available from your AI providers and selected in the Loop model allowlist. Loop resolves both organization-level and project-level providers, so when you open Loop within a project, models from that project’s providers are also available. If you see only Haiku and GPT-5-mini, only those models are allowed for Loop. Add Anthropic as a provider (with a valid API key), then select the desired Claude models in Loop settings, or allow additional models to expose them in the Loop dropdown.

​
What is happening

Loop reads from organization-level and project-level AI provider configurations, then applies the model allowlist from Loop settings. As an organization admin, you can trim the default model list to control token spend. If Anthropic is routed through Bedrock, Loop will not show friendly Claude names. Loop either shows Anthropic models when Anthropic is added directly, or shows Bedrock model identifiers if you keep Bedrock only. Organization-level providers and the Loop model allowlist are managed at the organization level by admins. Project-level providers can be added in a project’s AI provider settings, and their models become available to Loop when you open Loop within that project.

​
Fix or suggestion

​
Option 1: Add Anthropic as an organization provider (most common)

  1. Sign in as an organization administrator.
  2. Go to Settings > AI providers.
  3. Add a new provider entry for Anthropic. Enter the Anthropic API key.
  4. Go to Settings > Loop.
  5. Select the Claude models you want available to Loop (e.g., Sonnet, Opus).
  6. Save. Have users refresh the Loop chat window if models do not appear immediately.
Anthropic must be added directly (not via Bedrock) to show friendly Claude model names in Loop.

​
Option 2: Use Bedrock full model identifiers or add another provider

  1. If you prefer to keep Bedrock, confirm the specific Bedrock model identifiers are available from the org provider settings.
  2. Or add OpenAI (or another provider) at the org level.
  3. Go to Settings > Loop.
  4. Select the Bedrock model identifiers or GPT models you want available to Loop.
  5. Save and refresh Loop.
Bedrock model identifiers are not translated to Claude-style names. Use full Bedrock model IDs in Loop if you are not adding Anthropic directly.

​
How to confirm it worked

  • Open Loop and check the model dropdown for the newly enabled models or IDs.
  • Run a short test prompt selecting the model. The model should return a response and not immediately hit a provider error.

​
Notes

  • Enabling larger models increases token spend. Enable only the models you need to control cost.
  • If you hit provider max-token errors (for example with Bedrock/Opus), reduce context size or max tokens, use a smaller model, or limit that model’s availability in Loop.
  • Organization admins manage organization-level providers and the Loop model allowlist. Project-level providers can be added in a project’s AI provider settings. Changes propagate immediately; if a model does not appear, refresh the Loop chat window.