Open-source AI coding assistant

Your code.
Your model.
Your machine.

Launchbit is an open-source AI coding assistant. CLI-first, editor-native, and 100% local. Plug in any model you want — Anthropic, OpenAI, or your own self-hosted Llama. We never see your code.

Install View on GitHub

bash

$ npm i -g launchbit
$ cd ~/my-project
$ launchbit init
✓ Detected: TypeScript, Next.js 15
✓ Using model: claude-opus-4-8
>

Three principles

Built for developers who keep their context close.

Local-first

Your codebase never leaves your machine. The model API call is the only outbound network hop, and you control which endpoint.

Model-agnostic

Anthropic, OpenAI, Google, Mistral, or any OpenAI-compatible endpoint. Self-host Llama or Qwen via Ollama / vLLM and Launchbit just works.

Terminal-native

launchbit ask, launchbit fix, launchbit review. The CLI is the source of truth. Editor extensions wrap the same engine.

How it works

Three steps from clone to first patch.

Index your repo locally

Run launchbit init — embeddings are stored in .launchbit/, never uploaded.

Ask in natural language

launchbit ask "where is the auth middleware?"

Apply patches with diff preview

launchbit fix --interactive shows every change before it touches a file.

Supported models

Bring your own model. We hold none of the keys.

Anthropic ClaudeOpenAI GPTGoogle GeminiMistralOllama (local)vLLM (self-hosted)LiteLLMOpenRouter

Open source

Apache 2.0 licensed. github.com/launchbit/launchbit

Your code.Your model.Your machine.