
A local-first data engine that indexes everything you own and makes it searchable from one place, for humans and agents.
v3.0-beta · macOS · Linux · Windows

folder
gmail
apple-notes
chrome-bookmarks
chrome-history
obsidian
safari-history
github
slack
macos-calendar
notion
dropbox
google-drive
telegram
nas
linear
figma
spotify
Index any data source. Search everything locally. Keep it on your machine.
Full-text search (FTS5) and vector similarity (LanceDB) merged via Reciprocal Rank Fusion. Keyword precision and semantic understanding in a single query, fanned out across all repositories in parallel.
Script-based adapters in any language. A folder with a TOML manifest and any executable. Spacedrive writes JSON to stdin, the script writes JSONL to stdout. If it can print lines, it can be an adapter.
Each data source becomes a self-contained folder with its own SQLite database, vector index, and schema. Zip it, move it to another machine, drop it in the repos folder, done.
Everything runs on your machine. No cloud dependencies, no server processes. Single binary, embedded databases, local files. Your data never leaves your hardware.
Natively integrated with Spacebot as a Rust crate. No IPC, no protocol overhead. Your AI agent gets perfect recall of everything you own — emails, notes, bookmarks, browsing history, coding sessions.
Credentials stored in an encrypted secrets store — AES-256-GCM with Argon2id key derivation. Master key lives in the OS credential store, never on disk. Injected at sync time, never exposed.
Every record passes through safety screening, content classification, and quality scoring before it enters the search index. External data is scanned for prompt injection. Noise is detected and deprioritized. AI agents only see clean, ranked, trusted data.
Safety Screening
Prompt Guard 2 classifier catches injection attempts at ingest time. Quarantined content never enters the search index.
Content Classification
Every record is categorized and quality-scored. Bot output, spam, and auto-generated noise gets deprioritized automatically.
Ranked Indexing
Quality scores, safety verdicts, and trust tiers feed into search ranking. High-signal content surfaces first.
Index files by reference — metadata and content hashes, not copies. Search across every device, every drive, even when they're offline. Cross-device deduplication via BLAKE3 content hashing. Your files stay where they are. Spacedrive makes them all findable.
Reference-Only Indexing
Files stay on disk. Spacedrive stores paths, content hashes, extracted text, and rich metadata — never the files themselves.
Offline Volumes
Disconnected drives, NAS, and USB volumes stay in the index. Search results show availability. Reconnect and delta sync picks up where it left off.
Cross-Device Dedup
BLAKE3 content hashes identify identical files across devices. See exactly where duplicates live without moving anything.

An AI agent that thinks, executes, and responds concurrently. Built for teams and communities. Natively integrated with Spacedrive as a Rust crate — every search result your agent sees has been screened, classified, and ranked by the processing pipeline.
Channels handle conversations. Branches fork off to think. Workers execute tasks in parallel. Nothing blocks. Your agent works with clean, trusted data from Spacedrive — not raw dumps.
Your data is scattered across devices, apps, and services. Search is fragmented. AI tools can't see any of it unless you hand it to someone else's cloud.
You shouldn't need to surrender your data to make it useful.
Spacedrive started as a cross-platform file manager in 2021 — #1 on GitHub Trending, 37k stars, $2M seed. Two major versions proved the architecture works but the surface area of a full file manager was unsustainable. The most valuable part was never the file browser. It was the index. So we made that the product.
Local-first. Single binary. Never phones home. Your data stays where it is. Spacedrive just makes all of it findable from one place.