Use Open Interpreter with an uncensored LLM API

Open Interpreter is now a Rust program based on Codex, with a TOML config, not the older Python package that used LiteLLM. This page covers the new one. Custom providers are declared in config.toml.

Set it up

Sign in at wildwestapi.com/signup/ with the six-digit code sent to your email, open the dashboard and create an API key. It starts with sk-ww- and is shown in full once, so copy it somewhere safe.
Install it: on Windows run irm https://www.openinterpreter.com/install.ps1 | iex, and on macOS or Linux run curl -fsSL https://www.openinterpreter.com/install | sh.
Set WILDWEST_API_KEY in the environment you launch it from.
Open ~/.openinterpreter/config.toml and add a model_providers.wildwest table with base_url, env_key and wire_api set to chat.
Select the provider with model_provider and the model with model, then start it with interpreter.

config.toml

The config docs say user settings live in ~/.openinterpreter/config.toml, and that each provider is a [model_providers.<id>] table selected with model_provider. The keys are name, base_url, env_key (the name of the variable holding the key, not the key) and wire_api.

model_provider = "wildwest"
model = "outlaw-1"

[model_providers.wildwest]
name = "Wild West API"
base_url = "https://wildwestapi.com/v1"
env_key = "WILDWEST_API_KEY"
wire_api = "chat"

The docs say credentials should come from environment variables or the credential store, not inline tokens, and this layout follows that. A project can also have its own .openinterpreter/config.toml if the project is trusted.

Why wire_api is chat

The docs describe wire_api as the request format, with responses and chat as examples, and say to use chat for Chat Completions providers. Wild West documents the Chat Completions path, so set chat. For a one-off run without editing the file there is also interpreter --chat-completions, which overrides the format for that invocation only.

Choosing a model in the session

Inside the interface, /model switches providers and models. After setup you should see the Wild West provider there. Other ids are on the models page.

One limit to know

This tool runs commands and code on your machine. An uncensored model that is wrong or careless can do real damage to your files, so keep the approval prompts on and work in a throwaway folder or a container at first. The older Python interpreter.llm.api_base settings and the pip package do not apply to the new version.

FAQ

Which Open Interpreter is this?

The new Rust version installed from openinterpreter.com. The original Python project continues as a community fork, and its configuration is different.

Why does the key go in an environment variable?

The env_key setting names a variable to read, so the token never sits in the config file.

Use Aider with an uncensored LLM APIAider is a terminal pair programmer that edits files in your git repo. It reaches OpenAI-compatible servers through LiteLLM, so the setup is two environment variables and a model name with the openai/ prefix. Use Cline in VS Code with an uncensored LLM APICline is an autonomous coding agent for VS Code. It has a provider type called OpenAI Compatible, which is made for servers like this one. Everything is set in the settings panel, so there is no file to edit. Use Claude Code with an uncensored LLM APIClaude Code can be pointed at any gateway that serves the Anthropic Messages format, and Wild West serves it at /v1/messages. Read the limitation section first: Anthropic does not support this use.