Skip to main content
A project in Luminy is a local folder from your filesystem that you’ve connected to the agent. Once added, Luminy parses and indexes your code so the agent can read files, search symbols, navigate your directory tree, and make targeted edits — all without ever leaving your machine. Every conversation you start under a project inherits full awareness of that codebase, so you never have to paste in context manually.

​
Adding a project

1

Open the project picker

Click the + Projects button in the left sidebar. This opens your operating system’s native folder picker.
2

Select your folder

Navigate to and select the root directory of the codebase you want to connect. Luminy accepts any folder — a monorepo, a single package, or a personal scripts directory.
3

Wait for indexing to finish

Luminy immediately starts indexing the folder. A progress bar appears in the sidebar while the index is being built. The chat input is locked during this time — once indexing is complete the input unlocks and the agent is ready to work with your code.

​
What happens during indexing

When you add a folder, Luminy runs three steps in the background:

Parse source files

Tree-sitter tokenizes every source file it recognizes, extracting functions, classes, types, and other symbols.

Build a repo map

A compact structural map of the codebase is assembled from the parsed symbols and stored locally. This map lets the agent understand the shape of your code without reading every file from scratch on every request.

Generate a description

An AI-generated description of the project is produced automatically and attached to the project record. You’ll see this description in the sidebar.
Indexing is entirely local. Your source code is parsed and stored on your own machine — nothing is sent to any external server during the indexing process.

​
Supported languages

Tree-sitter indexing currently supports the following languages:
LanguageFile extensions
Python.py
JavaScript.js, .mjs, .cjs
TypeScript.ts, .tsx
Rust.rs
Go.go
Java.java
C++.cpp, .cc, .cxx, .hpp
C#.cs
Files in other languages are still readable and editable by the agent — they just don’t contribute symbols to the repo map.

​
How the agent uses your project

Once a project is indexed, the agent gains several codebase-aware capabilities:

File read & write

The agent can open, read, and edit any file in your project directory with precise line-level changes.

Grep & glob search

Pattern searches across the entire codebase let the agent find usages, imports, and references quickly.

Symbol lookup

The repo map gives the agent instant access to function signatures, class definitions, and type declarations without reading every file.

Repo map context

A structured overview of the project is injected into each conversation so the agent understands the shape of your codebase from the first message.

​
Managing projects

  • Switch projects — click any project in the sidebar to make it the active project. New sessions will be scoped to that project.
  • Delete a project — right-click a project in the sidebar and select Delete. This removes the project and its index from local storage. Your source files on disk are not affected.
If you add new files or make large structural changes to a codebase, you can re-index the project by removing it and adding it again. This rebuilds the repo map from scratch.