Skip to main content
Run /connect to add models from a supported provider, then use /model to select one. The choices shown depend on the providers and models available to your account.

​
GitHub Copilot

  1. Run /connect and choose GitHub Copilot.
  2. Open the displayed verification link and enter the code.
  3. Complete sign-in, then open /model and choose an available model.

​
ChatGPT Codex

  1. Run /connect and choose ChatGPT Pro/Plus.
  2. Choose browser or device-code sign-in and follow the prompts.
  3. Open /model and select an available model.
Use device-code sign-in when browser sign-in cannot start.

​
Manage providers

Return to /connect to disconnect or reconnect a provider. If a model disappears, check that the provider connection is still active and refresh it before selecting the model again. The Embedder’s own models switch lets you show only connected and local models. Choose a replacement in /model if the current model becomes unavailable. Some features may be unavailable with that switch off.

​
Choose a project default

Use /model to choose from the models available to your account. To pin a particular model when this project opens, create .embedder/models.json in the project root. For example, if GPT-5.5 (ChatGPT) is available through your ChatGPT connection:
.embedder/models.json
The value uses provider/exact-model-id. Use codex for the ChatGPT connection and github-copilot for GitHub Copilot. The model ID is the provider’s exact identifier, such as gpt-5.5, rather than a formatted display label such as GPT-5.5 (ChatGPT). For local models, use local/endpoint-name/model-id with the values from your local endpoint configuration. Check that the model is available in /model before pinning it. If you do not know its exact ID, select it through /model and leave main unset instead of guessing. A pin does not make an unavailable model accessible. Project model settings take precedence over ~/.embedder/models.json. Reopen the project after changing the file and check the selected model. If the expected model is not selected, check the JSON syntax, provider connection, and exact model ID. Set main to "inherit" to use ordinary model selection for this project even if your global file pins a model. Removing main instead falls back to the global setting, if one exists. For a model you host yourself, see Local models.
Last modified on September 12, 2026