Atlas Your files, connected the way you think.
Atlas maps the connections between your files — what depends on what, where things came from, and what breaks if something moves. All local, all private, all yours.
Your filesystem doesn't track connections.
Files don't exist in isolation — but your OS treats them like they do.
Move one file, break three projects. How would you know?
Why do you have three copies of this? Which one is the original?
What was that file called last week? Where did it go?
Three steps. Zero friction.
Watch
Point Atlas at the directories you care about. That's it.
Map
Atlas builds a graph of every connection — history, dependencies, shared origins, and forks between files.
Query
Ask questions about your files — what changed, what depends on what, what's affected if something moves.
Data designed to outlive the software.
Atlas stores everything in a single SQLite file on your machine. No server, no sync, no lock-in. Open it, query it, back it up — it's yours.
Append-only history
Nothing is overwritten or deleted. Every change is a new record. Your history is immutable and complete.
Content-addressed identity
Files are identified by what they contain, not where they live. Rename or move a file — Atlas knows it's the same thing.
Self-describing schema
The database documents itself. No proprietary format, no binary blob. Standard SQL, standard tools.
What Atlas sees that your OS can't.
Atlas identifies files by their content and maps the connections between them — giving you visibility into history, dependencies, and shared origins.
Dependency Mapping
See what files reference what. Find broken links, orphaned assets, and the cascade effects of a single change.
Fork Detection
Copy a file and Atlas recognizes the shared ancestor. See how copies diverge over time.
Scriptable
JSON output on every command. Pipe Atlas into your scripts, tools, and automations.
Content-Based Identity
Files are identified by a BLAKE3 fingerprint — not read or stored. Rename or move a file — Atlas knows it's the same entity.
Privacy by Default
Automatically skips credentials, keys, and secrets. No cloud. No accounts. Everything stays on your machine.
Set and Forget
Point Atlas at a directory and walk away. It runs in the background, building a complete map of your files — so the data is there when you need it.
See it in action.
Point Atlas at a directory. It handles the rest.
See how any file got here — every rename, move, and modification.
See the full chain of connections before you change anything.
See every copy and fork of a file — the full lineage, visualized.
A unified activity feed across everything Atlas tracks.
Find files that share the same content across projects and drives.
Built for one. Designed for many.
If your work lives across hundreds of files, Atlas gives you the visibility no other tool does. It starts as your personal file memory — and its data format is designed to work across machines and teammates without conflicts.
Sound Designers & Producers
Track which projects reference which samples. Move files without breaking sessions.
Web Developers
Trace CSS dependencies, find orphaned assets, and see the full impact of a change.
Game Developers
Update a texture, see every prefab and scene affected before you ship.
Filmmakers & Editors
Know which timelines use which clips. Find every version of a cut across drives.
Three layers. One system.
Atlas is designed as a modular stack — a core engine, domain-specific packs, and an interface for the tools you already use.
Atlas Core
The foundation. File watching, content hashing, history tracking, fork detection, and a built-in set of extractors for universal formats. Always on.
Atlas Packs
Domain-specific extractor bundles that teach Atlas how to read the relationships inside specialized file formats.
Atlas Connect
Exposes Atlas's knowledge graph to external tools and AI agents via MCP. Queryable by scripts, editors, and assistants.
Your filesystem, legible to intelligence.
AI agents are powerful — but they're blind to your filesystem. Every time an agent needs context about your files, it starts from scratch. Atlas Connect changes that.
Give agents memory
Atlas Connect exposes your file graph via MCP — the same protocol used by Claude, Cursor, and other AI tools. Agents can query dependencies, trace history, find duplicates, and map impact chains in real time — without reading your files or crawling your directories.
Metadata only. Never content.
Agents see names, paths, relationships, and timestamps. They never see what's inside your files. Every query is logged to an audit table you can inspect anytime. You always know what was asked and by whom.
The tools you already use, made smarter
Ask your coding assistant "what breaks if I rename this file?" and it can answer — because Atlas already mapped every dependency. Ask "where did this asset come from?" and get the full lineage. Atlas doesn't become intelligent. It becomes legible to intelligence.
- Search entities by name, path, or hash
- Query version history for any file
- Map dependencies and dependents
- Calculate impact trees
- Find broken references
- Discover orphaned assets
- List duplicate content
- View recent activity
Where we are. Where we're going.
Now
The core engine is built and running.
- File watching and BLAKE3 content hashing
- Automatic file tracking — renames, moves, modifications
- Fork detection and duplicate finding
- Full CLI with search, timeline, and tree views
- Background daemon with boot persistence
- Privacy-first ignore system
Next
Teaching Atlas to see the connections between files.
- Extractor framework and dependency graph
- Core Pack — HTML, CSS, Markdown, JSON, SVG extractors
- Dependency and impact queries
- Broken reference and orphan detection
- JSON output on every command
- Agent audit log and query controls
Future
Expanding what Atlas understands and how you interact with it.
- Domain packs for audio production, game development, and more
- AI agent integration via MCP — give your tools filesystem memory
- A desktop interface for visual exploration of your file graph
- Cross-machine awareness without cloud sync
Built with tools that last.
Rust
No runtime, no garbage collector. Fast, safe, and predictable.
SQLite
No server. Your data is one file. Back it up, query it, own it.
BLAKE3
Cryptographic hashing at gigabytes per second. Fast, secure, content-addressable.