Chat with a local AI model — right inside VS Code, fully offline, powered by Ollama. No API keys, no subscription, no code ever leaving your machine.
No servers, no accounts: pick a model, download it once, and keep coding with it even with the internet turned off.
Features
- Guided setup, no terminal required — the sidebar detects whether Ollama is running, offers one-click install (Linux, and macOS with Homebrew), and shows a live progress bar while a model downloads.
- 39 curated models across 3 categories — Code, General, and Vision — each with a short description and approximate download size.
- Model management in the sidebar — switch between installed models with one click, delete ones you no longer need, or add any model from the Ollama library by name.
- A real chat panel, opening in a split to the right of the editor — streamed replies, a working Stop button, and code blocks rendered with your editor's own theme colors (with Copy / Insert into editor buttons).
- Copy or edit any message — copy just the prose of a reply, or edit and resend one of your own messages.
- File & image attachments — attach a text/code file for context with any model; attach an image only when a vision-capable model is active.
- Ask about selected code straight from the editor's right-click menu.
Usage
- Click the Local Full AI icon in the activity bar.
- If prompted, install Ollama (automatically, or via the official installer).
- Pick a model from the list — Code, General, or Vision — and wait for it to download.
- Click "Open chat" to start talking to it, or select some code and choose "Ask about selected code" from the right-click menu.
- Once a model is downloaded, everything works with no internet connection at all.
License
MIT