> ## Documentation Index
> Fetch the complete documentation index at: https://docs.embedder.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect subscriptions

> Connect GitHub Copilot or ChatGPT Codex so Embedder can use those subscriptions for inference.

Use `/connect` to add models from a GitHub Copilot or ChatGPT subscription. Connected models appear in `/model`.

<Note>
  Connecting a subscription only adds models. Your Embedder account, team, project, and tools do not change.
</Note>

## GitHub Copilot

<Steps>
  <Step title="Choose GitHub Copilot">
    Run `/connect` and choose **GitHub Copilot**.
  </Step>

  <Step title="Sign in to GitHub">
    Open the verification URL, enter the displayed code, and approve access.
  </Step>

  <Step title="Choose a model">
    Open `/model` and select one of the models available to your Copilot account.
  </Step>
</Steps>

## ChatGPT Codex

<Steps>
  <Step title="Choose ChatGPT Pro/Plus">
    Run `/connect` and choose **ChatGPT Pro/Plus**.
  </Step>

  <Step title="Sign in">
    Choose browser sign-in or device-code sign-in and follow the prompts. Use the device-code option if browser sign-in is unavailable.
  </Step>

  <Step title="Choose a model">
    Open `/model` and select one of the models available to your ChatGPT account.
  </Step>
</Steps>

## Manage connections

Use `/connect` to disconnect or reconnect a provider. Credentials are stored on this machine at:

```text theme={"system"}
~/.embedder/connected-providers.json
```

<Warning>
  `connected-providers.json` is local JSON storage, not an operating-system keychain. Restrict access to the Embedder application directory.
</Warning>

## Pin a model to a project

Copilot and Codex use the `<provider>/<model>` format:

```json .embedder/models.json theme={"system"}
{
  "main": "github-copilot/claude-sonnet-4",
  "fast": "codex/gpt-5.6-sol"
}
```

Project settings override `~/.embedder/models.json`. A valid `main` pin is selected when the project opens.

## Hide Embedder-hosted models

In `/connect`, turn off **Embedder's own models** to show only subscription and local models. If the current model is removed, choose another model in `/model`.

Features that rely on Embedder-hosted inference are unavailable while this switch is off.

## Troubleshooting

<AccordionGroup>
  <Accordion title="Codex browser login cannot start">
    Port `1455` is already in use. Stop the process using it or choose the Codex device-code method.
  </Accordion>

  <Accordion title="A model is missing">
    GitHub and OpenAI control which models each account can use. Check your subscription and organization policy, then reconnect the provider to refresh the list.
  </Accordion>
</AccordionGroup>

<CardGroup cols={2}>
  <Card title="Local models" icon="server" href="/integrations/local-models">
    Connect Ollama, LM Studio, vLLM, llama.cpp, or another OpenAI-compatible endpoint.
  </Card>
</CardGroup>