A semantic filesystem, built in the open, that you can fully own.

Why we're building this

Companies have decades of files. They live in laptops, in S3 buckets, in SharePoint sites, in Jira projects, in a Drive folder someone created in 2018. Knowing what's in them — and getting an AI agent to use that context — used to require either expensive SaaS or weeks of glue code.

We thought you should be able to point one binary at all of that, get a ranked search in 100 ms, and hand the same engine to your AI agent through the Model Context Protocol. No vendor in the path. No file contents leaving the machine. So we built it.

What's in the box today

• Seven storage backends: Local, S3, GCS, Azure Blob, SharePoint, Google Drive, Jira.
• 35+ parsed file types: code, prose, spreadsheets, presentations, configs.
• Hybrid search fusing SQLite FTS5 BM25, sqlite-vec cosine similarity, and metadata.
• Native Model Context Protocol server on port 8081.
• REST API on port 8080 with an OpenAPI schema.
• Desktop UI (Wails) with native installers for macOS, Linux, Windows.
• FUSE / WinFsp mount that exposes the index as a read-only filesystem.
• OS-level integration: Spotlight importer, GNOME Shell SearchProvider, Windows Search IFilter.

Project status

• Version: 0.2.1
• License: MIT
• Stage: active development; production-ready for single-user / small-team deployments.
• Repository: github.com/neul-labs/stratafs
• Issues: github.com/neul-labs/stratafs/issues
• Discussions: github.com/neul-labs/stratafs/discussions

What's next

In flight or imminent:

• Enterprise RBAC + source-level permissions.
• Streaming search results (chunked Transfer-Encoding).
• SQLCipher-backed at-rest encryption for source databases.
• Cross-source ranking signals (more weight to internal docs over public mirrors).

Further out: distributed read replicas, OCR for images and scanned PDFs, multimodal embeddings.

Who's behind it

StrataFS is a project of Neul Labs, with development led by Dipankar Sarkar. The project is fully open source under the MIT license. Contributions, bug reports, and design discussions are welcome through the GitHub repository.

The name

"Strata" — layers of sediment. "FS" — filesystem. The picture is of passive storage as geological strata, and StrataFS as the survey that reveals what's actually in each layer. Less metaphorically: every file type is a stratum we know how to read.