Spacedrive
Spacedrive is an open source cross-platform file explorer, powered by a virtual distributed filesystem written in Rust.
Install / Use
/learn @spacedriveapp/SpacedriveREADME
Spacedrive is an open source cross-platform file manager, powered by a virtual distributed filesystem (VDFS) written in Rust.
Organize files across multiple devices, clouds, and platforms from a single interface. Tag once, access everywhere. Never lose track of where your files are.
[!IMPORTANT] v2.0.0-alpha.1 Released: December 26, 2025
This is Spacedrive v2—a complete ground-up rewrite. After development of the original alpha version stopped in January this year, I rebuilt Spacedrive from scratch with the hard lessons learned.
Current status: Alpha release for macOS and Linux. Windows support coming in alpha.2. Mobile apps (iOS/Android) coming soon.
Download Release · Visit v2.spacedrive.com for complete documentation and guides.
If you're looking for the previous version, see the v1 branch.
The Problem
Computing was designed for a single-device world. The file managers we use today—Finder, Explorer, Files—were built when your data lived in one place: the computer in front of you.
The shift to multi-device computing forced us into cloud ecosystems. Want your files everywhere? Upload them to someone else's servers. The convenience came at a cost: data ownership. This wasn't accidental—centralization was the path of least resistance for solving multi-device sync.
Now AI is accelerating this trend. Cloud services offer intelligent file analysis and semantic search, but only if you upload your data to their infrastructure. As we generate more data and AI becomes more capable, we're giving away more and more to access basic computing conveniences.
The current system isn't built for a world where:
- You own multiple devices with underutilized compute and storage
- Local AI models are becoming competitive with cloud alternatives
- Privacy and data sovereignty matter
- You shouldn't have to choose between convenience and control
The Vision
Spacedrive is infrastructure for the next era of computing. It's an architecture designed for multi-device environments from the ground up—not cloud services retrofitted with offline support, but local-first sync that scales to the cloud when you want it.
As local AI models improve, Spacedrive becomes the fabric that enables the same insights cloud services offer today, but running on hardware you already own, on data that never leaves your control. This is a long-term project correcting computing's trajectory toward centralization.
The file explorer interface is deliberate. Everyone understands it. It's seen the least innovation in decades. And it has the most potential when you bake distributed computing, content awareness, and local AI into something universally familiar.
How It Works
Spacedrive treats files as first-class objects with content identity, not paths. A photo on your laptop and the same photo on your NAS are recognized as one piece of content. This enables:
- Content-aware deduplication - Track redundancy across all devices
- Semantic search - Find files in under 100ms across millions of entries
- Transactional operations - Preview conflicts, space savings, and outcomes before execution
- Peer-to-peer sync - No servers, no consensus protocols, no single point of failure
- Offline-first - Full functionality without internet, syncs when devices reconnect
Files stay where they are. Spacedrive just makes them universally addressable with rich metadata and cross-device intelligence.
Architecture
Spacedrive is built on four core principles:
1. Virtual Distributed Filesystem (VDFS)
Files and folders become first-class objects with rich metadata, independent of their physical location. Every file gets a universal address (SdPath) that works across devices. Content-aware addressing means you can reference files by what they contain, not just where they live.
2. Content Identity System
Adaptive hashing (BLAKE3 with strategic sampling for large files) creates a unique fingerprint for every piece of content. This enables:
- Deduplication: Recognize identical files across devices
- Redundancy tracking: Know where your backups are
- Content-based operations: "Copy this file from wherever it's available"
3. Transactional Actions
Every file operation can be previewed before execution. See exactly what will happen—space savings, conflicts, estimated time—then approve or cancel. Operations become durable jobs that survive network interruptions and device restarts.
4. Leaderless Sync
Peer-to-peer synchronization without central coordinators. Device-specific data (your filesystem index) uses state replication. Shared metadata (tags, ratings) uses a lightweight HLC-ordered log with deterministic conflict resolution. No leader election, no single point of failure.
Core Features
| Feature | Description | | ----------------------- | ---------------------------------------------------------------------------- | | Cross-Platform | macOS, Windows, Linux, iOS, Android | | Multi-Device Index | Unified view of files across all your devices | | Content Addressing | Find optimal file copies automatically (local-first, then LAN, then cloud) | | Smart Deduplication | Identify identical files regardless of name or location | | Cloud Integration | Index S3, Google Drive, Dropbox as first-class volumes | | P2P Networking | Direct device connections with automatic NAT traversal (Iroh + QUIC) | | Semantic Tags | Graph-based tagging with hierarchies, aliases, and contextual disambiguation | | Action Preview | Simulate any operation before execution | | Offline-First | Full functionality without internet, syncs when devices reconnect | | Local Backup | P2P backup between your own devices (iOS photo backup available now) | | Extension System | WASM-based plugins for domain-specific functionality |
Tech Stack
Core
- Rust - Entire VDFS implementation (~183k lines)
- Tokio - Async runtime
- SQLite + SeaORM - Local-first database with type-safe ORM queries
- Iroh - P2P networking with QUIC transport, hole-punching, and local discovery
- BLAKE3 - Fast cryptographic hashing for content identity
- Wasmer - Sandboxed WASM extension runtime
- Axum - HTTP/GraphQL server for web and API access
- OpenDAL - Unified cloud storage abstraction (S3, Google Drive, OneDrive, Dropbox, Azure Blob, GCS)
- Specta - Auto-generated TypeScript and Swift types from Rust
Cryptography & Security
- Ed25519 / X25519 - Signatures and key exchange
- ChaCha20-Poly1305 / AES-GCM - Authenticated encryption
- Argon2 - Password hashing
- BIP39 - Mnemonic phrase support for key backup
- redb - Encrypted key-value store for credentials
Media Processing
- FFmpeg (via custom
sd-ffmpegcrate) - Video thumbnails, audio extraction - libheif - HEIF/HEIC image support
- Pdfium - PDF rendering
- Whisper - On-device speech recognition (Metal-accelerated on Apple platforms)
- Blurhash - Compact image placeholders
Interface (shared across web and desktop)
- React 19 - UI framework
- Vite - Build tooling
- TypeScript - Type-safe frontend code
- TanStack Query - Server state management
- Zustand - Client state management
- Radix UI - Accessible headless components
- Tailwind CSS - Utility-first styling
- Framer Motion - Animations
- React Hook Form + Zod - Form management and validation
- Three.js / React Three Fiber - 3D visualization
- dnd-kit - Drag and drop
- TanStack Virtual / TanStack Table - Virtualized lists and tables
Desktop
- Tauri 2 - Cross-platform desktop shell (macOS, Linux, Windows)
Mobile (React Native)
- React Native 0.81 + Expo - Cross-platform mobile framework
- Expo Router - File-based routing
- NativeWind - Tailwind CSS for React Native
- React Navigation - Native navigation stack
- Reanimated - Native-thread animations
- sd-mobile-core - Rust core bridge via FFI
Architecture Patterns
- Event-driven design with centralized EventBus
- CQRS: Actions (mutations) and Queries (reads) with preview-commit-verify
- Durable jobs with MessagePack serialization and checkpointing
- Domain-separated sync with clear data ownershi
