LocalPilot IDE
A fully local, browser-based AI coding IDE with Copilot-style completions, powered entirely by a local Ollama model.
An AI IDE that never phones home
LocalPilot is a browser-based AI coding IDE built on Monaco Editor — the same editor VS Code uses — with GitHub Copilot-style ghost-text completions and an AI agent that can generate, edit, and debug code through chat.
Every request is routed to a local Ollama model instead of a hosted API, so the whole thing — completions, chat, and file operations — runs on one machine with no data leaving it. It's still an experimental project, but a 7B coding model over FastAPI turned out to be enough for real inline suggestions and multi-file scaffolding.
What it does
- File explorer, Monaco code editor, and AI chat side by side
- Same editor engine that powers VS Code
- GitHub Copilot-style ghost-text completions as you type
- AI agent generates, edits, debugs, and explains code via chat
- Ask the AI to scaffold an entire multi-file project from scratch
- Create, rename, and delete files and folders in the browser
- ZIP upload to import an existing project, ZIP export to download results
- Every request routes to a local Ollama model — no data leaves the machine
- No accounts, API keys, or cloud calls required
How a completion happens
You type in the Monaco editor, or open the AI chat panel with a request.
The FastAPI backend builds a context window from the open file(s) and sends it to Ollama.
Ollama (Qwen2.5-Coder 7B) streams back an inline completion or a chat response.
Ghost text renders directly in the editor, or the AI agent applies file edits and creates new files through the same backend that powers the file explorer.
Status: experimental. It works end-to-end on a 7B model over a local GPU; larger models (e.g. 70B) haven't been tested yet. Runs GPU-accelerated via docker-compose.yml, or CPU-only via the docker-compose.cpu.yml override.
Tech Stack
Check out the code
Full source, setup instructions, and architecture notes are on GitHub.