Spacedrive

Spacedrive is an open-source, cross-platform data platform that replaces traditional path-based file management with a content-addressed Virtual Distributed File System (VDFS). Every file gets a BLAKE3 content hash that serves as its universal identity across all your machines, enabling true deduplication and cross-device awareness without moving or copying data. Whether a file lives on a local SSD, an external NAS, a cloud bucket, or an offline device, Spacedrive tracks it as a single object with a unified address.

Version 2.0 represents a ground-up rewrite in Rust that adds leaderless peer-to-peer sync via Iroh/QUIC, nine distinct file views (grid, list, columns, media, size, recents, search, knowledge, and splat), and a script-based adapter system that lets you index arbitrary data sources — Gmail, Obsidian vaults, Slack messages, Chrome history, GitHub stars — as first-class volumes alongside your files. All indexing and search runs entirely on your machine with no cloud dependency.

Spacedrive also introduces a novel AI safety layer: before any indexed content is exposed to AI agents, it passes through a local Prompt Guard 2 classifier that detects prompt injection attacks in emails, documents, and messages. The companion Spacebot agent runtime routes all AI operations through Spacedrive’s structured permission and access model instead of granting agents raw shell access, making it the first local file manager designed from the ground up for safe AI integration.

The application targets macOS, Windows, Linux, iOS, and Android through a shared Rust core exposed via Tauri on desktop and React Native on mobile, with an optional headless server mode and a full-featured CLI for scripting and automation workflows.

What You Get

• Virtual Distributed File System (VDFS) — Every file is indexed by its BLAKE3 content hash and assigned a universal SdPath address that works across all your devices and storage backends, enabling deduplication and cross-device awareness without copying data.
• Leaderless P2P Sync — Devices connect directly via Iroh/QUIC with automatic NAT traversal, local network discovery, and relay fallback. Metadata syncs between devices using an HLC-ordered log with deterministic conflict resolution — no central server or coordinator required.
• Script-Based Adapter System — Index any external data source by writing a folder with an adapter.toml manifest and a script in any language that reads stdin and writes JSONL. Eleven adapters ship out of the box: Gmail, Apple Notes, Chrome/Safari history, Obsidian, Slack, GitHub, macOS Contacts, Calendar, and OpenCode.
• Nine File Views — Switch between grid, list, columns, media, size, recents, search, knowledge, and splat views. QuickPreview handles video, audio, code, documents, 3D models, and images inline without leaving the app.
• AI Safety Screening — All indexed content passes through a local Prompt Guard 2 classifier before being exposed to AI agents, detecting prompt injection in emails and documents. A trust tier system quarantines flagged records and gives agents trust metadata alongside search results.
• Spacebot Agent Integration — An optional open-source AI agent runtime that pairs with Spacedrive as its home node. Agents access files, run shell commands, and operate across your entire device fleet through Spacedrive’s permission layer — no raw shell access granted.
• Cloud Volumes — S3, Google Drive, Dropbox, OneDrive, Azure Blob, and Google Cloud Storage appear as first-class volumes alongside local storage. Cross-storage operations (local to cloud, cloud to cloud) work natively.
• Transactional File Operations — Every operation (copy, move, delete, sync) can be previewed before execution to see space savings, conflicts, and estimated time. Operations run as durable, resumable jobs that survive device restarts and network interruptions.
• At-Rest and In-Transit Encryption — Libraries can be encrypted on disk via SQLCipher. All P2P traffic is encrypted via QUIC/TLS with Ed25519/X25519 key pairs. Credentials for cloud services are stored in OS-native credential stores (Keychain, Windows Credential Manager).
• Headless Server and CLI — Run Spacedrive as a server for NAS or home lab deployments. The full-featured sd-cli enables library management, location operations, and job scripting without a GUI.

Common Use Cases

• Cross-device file search — A photographer searches for RAW files across a MacBook, external RAID, and NAS without syncing copies to any central location, relying on Spacedrive’s content-hash index to surface results from offline devices too.
• Personal knowledge base indexing — A researcher connects Obsidian vaults, Gmail threads, Apple Notes, and Safari history as adapter sources, then uses semantic search (LanceDB + FastEmbed) to find related content across years of notes and archived emails in one query.
• Safe AI agent access to local files — A developer runs Spacebot pointed at their Spacedrive library so an AI coding assistant can read project files and documentation without being granted raw filesystem access, with all operations audited through the permission layer.
• Home lab file server — A self-hoster runs Spacedrive in headless server mode on a Raspberry Pi, pairs it with their laptop and phone over P2P, and gets a private, encrypted file sync and search layer without subscribing to any cloud service.
• Cloud storage consolidation — A team indexes S3 buckets, Google Drive, and Dropbox as volumes in Spacedrive to search and move files across cloud providers from one interface, with deduplication identifying identical files stored redundantly across services.
• Archival and redundancy tracking — A sysadmin uses Spacedrive’s content identity system to find duplicate files across backup drives and cloud buckets, freeing storage by removing redundant copies while keeping the original indexed.

Under The Hood

Architecture Spacedrive’s core is a single Rust crate structured around CQRS and domain-driven design with sharply separated layers. The domain layer defines the VDFS primitives — files as content-addressed objects with SdPath universal addresses, libraries as isolated namespaces, volumes as pluggable storage backends — without any infrastructure dependencies. Every user-initiated operation is encoded as a typed CoreAction or LibraryAction trait implementor with explicit input, output, and validation contracts; an ActionManager auto-registers all operations at startup, and a unified ApiDispatcher routes calls identically whether they arrive from the Tauri desktop IPC, the CLI daemon over Unix sockets, or the mobile FFI layer. Dependency injection flows through an Arc-wrapped CoreContext holding managers for devices, libraries, volumes, and the event bus, making components independently testable. The P2P sync subsystem is fully decoupled from library internals, communicating through typed events and an HLC-ordered leaderless log with deterministic conflict resolution — no central coordinator is needed at any point.

Tech Stack The backend runs on Tokio with SQLite accessed via SeaORM and sqlx, SQLCipher for at-rest encryption, and LanceDB with FastEmbed for vector-based semantic search. P2P networking uses Iroh over QUIC with Ed25519/X25519 key pairs and ChaCha20-Poly1305 transport encryption, automatic NAT traversal, local discovery, and relay fallback. BLAKE3 content hashing provides the content identity layer, while OpenDAL abstracts cloud storage across S3, Google Drive, Dropbox, OneDrive, Azure, and GCS behind a uniform API. Media processing uses FFmpeg, libheif, Pdfium, and Whisper for transcription. The desktop app runs on Tauri 2 with React 19, Vite, TanStack Query, and Tailwind CSS v4; mobile targets iOS and Android via React Native and Expo. Specta auto-generates TypeScript client types from Rust action definitions, enforcing end-to-end type safety across the Rust/TypeScript boundary. Build orchestration uses Bun and Turbo across the monorepo.

Code Quality The integration test suite is comprehensive, with 56 test files covering the full operational surface: file copy and move, cross-device sync, device pairing over proxy and relay, WASM extension execution, volume detection and fingerprint stability, TypeScript bridge round-trips, and backfill race conditions under concurrent writes. Tests use an in-process Core with realistic mock volumes and async network simulators. Error handling is explicit throughout — the action system returns typed ValidationResult and Receipt enums rather than booleans, and Rust’s type system enforces invariants at compile time via serde and Specta-generated types. Clippy and rustfmt run in CI; the frontend enforces Prettier with import sorting. One pragmatic signal of the alpha phase is a crate-level #![allow(warnings)], indicating some technical debt being actively worked through during the v2 rewrite.

What Makes It Unique Spacedrive introduces several primitives with no direct equivalent in other local file tools. The script-based adapter system — a folder with an adapter.toml and a script in any language that reads stdin and writes JSONL — lets any developer turn any data source into a first-class searchable volume without a proprietary SDK. The AI safety screening pipeline using a local Prompt Guard 2 classifier to detect prompt injection in emails and documents before they enter the search index is the first such primitive in a local file manager. The Spacebot integration routes all AI agent operations through a structured permission and audit layer instead of granting raw shell access, establishing a new model for local-first AI computing. Leaderless P2P sync using HLC-ordered logs with deterministic conflict resolution — combined with VDFS content addressing — enables true multi-device file management without any central server, coordinator, or cloud subscription.

Self-Hosting

Spacedrive is licensed under the Functional Source License version 1.1 with an Apache 2.0 future grant (FSL-1.1-ALv2). In plain terms, this means you can use, copy, modify, and redistribute the software freely for personal use, education, research, and professional services — as long as you are not offering a competing commercial product or service. Specific restrictions prohibit running Spacedrive as a managed cloud service, offering commercial Spacedrive hosting to third parties, or building commercial AI agent platforms on top of it. On the second anniversary of each release, the code for that version converts automatically to the Apache License 2.0, becoming fully permissive. For most self-hosters and developers building internal tools, FSL-1.1-ALv2 is permissive enough to use freely — the restrictions only bite if you intend to resell Spacedrive-as-a-service commercially.

Running Spacedrive yourself is non-trivial at this stage. The project is in public alpha (v2.0.0-alpha.2 as of early 2026), and the setup requires Rust 1.81+, Bun 1.3+, just, and Python 3.9+ for adapters. The desktop app compiles a full Rust core, native media processing libraries (FFmpeg, libheif, Pdfium, Whisper), and a Tauri frontend before producing a runnable binary. For headless server deployments on a NAS or home lab, you are responsible for managing the daemon process, library backups, SQLite database maintenance, and P2P relay configuration if your devices cannot reach each other directly via local network or NAT traversal. There is no official Docker image or one-click installer for the server edition at this time. The operational burden is consistent with other Rust-native, self-hosted tools at the alpha stage — expect rough edges and plan for occasional manual intervention during upgrades.

Spacedrive does not currently offer a paid hosted tier or commercial cloud service. The project is venture-backed and plans to monetize through optional cloud relay, backup, and managed services in the future, but those products do not exist yet. For self-hosters, this means there are no SLAs, no managed upgrade paths, no enterprise support contracts, and no HA configuration guides. What you gain is complete data sovereignty — no telemetry, no third-party servers in the data path, no subscription required — and the ability to run your own relay servers if needed. Compared to Dropbox or Google Drive, you trade away seamless mobile sync, automatic versioning, and 24/7 support in exchange for keeping every byte under your own control.