Models & providers

Connect a provider

Open /login, choose a provider, and follow its authentication flow. Use /models to switch models after connecting.

e stores credentials in ~/.e/auth.json. Supported providers can also read their conventional environment variables, such as OPENAI_API_KEY or ANTHROPIC_API_KEY. Stored credentials take precedence.

e does not include a model subscription. Your provider's billing and usage limits apply.

Run a local model

Ollama at localhost:11434 and LM Studio at localhost:1234 do not need a stored credential. Start the server and load a model; e discovers its models when the endpoint responds.

A model needs tool-calling support to use e's file and shell tools. Configure its context window to match the model and server.

Add a custom endpoint

Create ~/.e/models.json. This example describes a local OpenAI-compatible server; replace the endpoint, model ID, and context window with your server's values.

{
  "providers": {
    "local": {
      "base_url": "http://localhost:8080/v1",
      "api": "openai-completions",
      "context_window": 64000,
      "supports_tools": true,
      "models": [
        "your-model-id"
      ]
    }
  }
}

The api field selects the request format. Supported formats include OpenAI Completions and Responses, Anthropic Messages, and Google Generative AI. A new provider requires an explicit base_url.

Model discovery

e refreshes the model catalog at launch, after sign-in, and when you open /models. An entry in models.json with the same provider and model ID overrides the built-in entry.

Use /scoped-models to choose a shortlist. Ctrl + P cycles through it. For the complete configuration format, run e docs models.