# twinny

An open-source AI coding assistant for VS Code that provides code completion, inline edits, chat, code review and commit messages against a model server the user controls, with no telemetry or sign-in.

twinny is an AI coding assistant for Visual Studio Code designed to keep all model traffic inside a user's own network. It offers fill-in-the-middle code completion streamed as ghost text, inline edits reviewed as a diff and accepted or rejected per change, chat that can attach files, symbols, the problems panel, a git diff or a workspace search, a hybrid keyword-and-vector workspace index reranked before reaching the model, code review of a working tree, a branch or a pull request, commit-message generation from a staged diff, and a terminal helper that writes and explains shell commands.

It can run against a model server on the user's own machine, another of their devices, a hosted API, or one gateway shared by a whole team, using Ollama, LM Studio or llama.cpp for local models. A companion server command sets up a shared team gateway with per-person keys, usage tracking and an admin page.

It is free, open source, licensed under MIT, and has no telemetry or required sign-in.

Website: https://twinny.dev