twinny: AI coding assistant for Visual Studio Code

twinny is an AI coding assistant for Visual Studio Code that works with a model server you choose: local, another device, a hosted API or a team gateway.

Verified2026-11-01

Core facts

Read by a script from the official API, not rewritten by a model.

Repository: https://github.com/twinnydotdev/twinny
License: MIT
Languages: TypeScript
Maintainer: twinnydotdev
Homepage: https://twinny.dev
Verified: 2026-11-01

What it is

The project describes twinny as an AI coding assistant for Visual Studio Code that stays inside your network; the model server is one you choose, on your machine, on another of your devices, on a hosted API, or on one gateway your whole team shares, and a hosted API setup is outside that stays-inside-your-network scope. twinny is free, open source and MIT licensed, with no telemetry and no sign-in. The project says everything runs against a server you control; that applies to self-managed backends such as your machine, your other devices and the team gateway, not to a hosted API setup. twinny is written and maintained by one person since 2023.

Main features

Code completion takes its context from open files, imports, the language server and recent edits.
Inline edit lets you describe a change with Ctrl+I and review it as a diff in the editor, and changes are accepted or rejected per hunk.
In chat, typing @ attaches files, symbols, the problems panel, the git diff, the terminal, or a search of the workspace index, and conversations are kept.
Agent mode (experimental) uses a switch in the chat or Shift+Tab to let the model read, search and edit files and run commands in the workspace, with each step shown in the reply.
The workspace index provides hybrid keyword and vector search over the workspace, reranked before it reaches the prompt and updated on save.
Code review covers the working tree, a branch against its base, or a GitHub pull request.

Getting started

twinny is installed from the Marketplace, with VS Code 1.93 or newer.
At first start, twinny finds Ollama, LM Studio and llama.cpp on their usual ports and asks which models to use.
twinny-server runs on the machine with the models and serves chat, completion and embeddings to developers' VS Code.
From the admin page you set the team's default models, send each developer an invite link that opens VS Code and connects them, and see usage per person and per model.

From the README · sh

npx twinny-server quickstart

Who it is for

In agent mode, commands ask first unless you say otherwise. Team policy needs a licence and covers team-only providers, locked defaults, routing rules that keep a workspace on local backends, and a team system prompt. Prompts in twinny-server go to your gateway and your backend, nowhere else.

twinny-server issues a key per developer, stored as a hash and revoked live. A developer flips Share this computer, and their local server serves the team through the gateway. Operations in twinny-server include a hash-chained audit log of every admin change, read-only admin keys, Prometheus metrics, costs per developer when a model has a price, a request queue for a shared GPU, Docker and a Helm chart.

The hosted API providers are OpenAI, Anthropic, Mistral (Codestral for completion), DeepSeek, OpenRouter, Gemini, Groq, Cohere and Perplexity. Providers can be mixed, for example a local model for completion and a hosted one for chat. twinny is actively developed and provided as is.