Authentication

Open Interpreter is provider-agnostic. The first run asks how you want to authenticate, and /model lets you change that later.

Start the TUI and choose ChatGPT sign-in:

interpreter

This opens a browser-based login flow and stores refreshable credentials in the configured credential store.

API Keys

API keys are the best fit for CI, headless machines, and explicit provider billing.

export OPENAI_API_KEY=sk-...
interpreter

Other providers use their own environment variables, such as ANTHROPIC_API_KEY, or the variable configured by their provider entry.

Local Providers

Use local runners when you want model traffic to stay on your machine:

Provider	Notes
Ollama	Start Ollama and choose it from `/model`, or use `--oss`.
LM Studio	Start the local server and choose LM Studio from `/model`.

interpreter --oss "summarize this repo with my local model"

Compatible Providers

Add an OpenAI-compatible provider in config:

model_provider = "acme"
model = "acme-coder"

[model_providers.acme]
name = "Acme"
base_url = "https://api.acme.example/v1"
env_key = "ACME_API_KEY"
wire_api = "responses"

Then:

export ACME_API_KEY=...
interpreter

Credential Storage

cli_auth_credentials_store = "auto" # "auto" | "keyring" | "file"

Open Interpreter stores user state under ~/.openinterpreter/. Treat auth.json like a password if file storage is used.

Sign Out

Inside the TUI:

/logout

Or from a compatible login surface, use the logout command when available.