# StrataFS

> A semantic filesystem that turns Local, S3, GCS, Azure, SharePoint, Google Drive and Jira into an AI-searchable knowledge base.

StrataFS is an open-source semantic filesystem for the AI era. Index local and cloud storage, search with hybrid FTS5 + vector retrieval, and expose your data to AI agents via a native Model Context Protocol (MCP) server — no SaaS, no API calls, no lock-in.

## Key URLs

- [Home](https://stratafs.neullabs.com/): Marketing home with hero, features, live demo, and FAQ.
- [Architecture](https://stratafs.neullabs.com/architecture): How the pipeline, storage, queue, parsers, embeddings, database, and APIs fit together.
- [Search engine](https://stratafs.neullabs.com/search): Hybrid BM25 + vector + metadata ranking, explained with the SQL query shape.
- [MCP server](https://stratafs.neullabs.com/mcp): Native Model Context Protocol integration for AI agents.
- [Storage backends](https://stratafs.neullabs.com/storage): Local, S3, GCS, Azure, SharePoint, Drive, and Jira with config examples.
- [Install](https://stratafs.neullabs.com/install): npm, pip, curl, Homebrew, Docker, and source.
- [For developers](https://stratafs.neullabs.com/use-cases/developers): Codebase search that beats grep on intent queries.
- [For AI engineers](https://stratafs.neullabs.com/use-cases/ai-engineers): Retrieval that knows it's talking to an LLM.
- [For enterprise](https://stratafs.neullabs.com/use-cases/enterprise): SharePoint, Drive, Jira indexing without SaaS lock-in.
- [Compare](https://stratafs.neullabs.com/compare): vs. Elasticsearch, Pinecone, grep.
- [FAQ](https://stratafs.neullabs.com/faq): Common questions about install, search, MCP, and operations.
- [About](https://stratafs.neullabs.com/about): Project goals, status, roadmap.
- [GitHub](https://github.com/neul-labs/stratafs): Source code, issues, releases.
- [Releases](https://github.com/neul-labs/stratafs/releases): Versioned binaries for macOS, Linux, Windows.

## Articles

- [StrataFS vs. Elasticsearch for semantic code search — when each makes sense](https://stratafs.neullabs.com/blog/stratafs-vs-elasticsearch-for-developers): Elasticsearch is the workhorse of full-text search. StrataFS is the new workhorse of embedded semantic search. Here's an honest comparison for developer-focused use cases.
- [Claude Desktop + StrataFS: a five-minute MCP setup for your codebase](https://stratafs.neullabs.com/blog/claude-desktop-mcp-stratafs-setup): Wire StrataFS into Claude Desktop in five minutes. Hybrid search across your code and docs becomes a tool Claude knows how to call. Step-by-step, with the config and the troubleshooting.
- [Index S3, GCS, and Azure Blob buckets with semantic search](https://stratafs.neullabs.com/blog/index-s3-gcs-azure-with-stratafs): Indexing cloud storage used to mean pipelines, copies, and a separate search service. StrataFS reads buckets in place with read-only credentials and exposes a hybrid search across all of them.
- [Self-hosted semantic search: local AI indexing without SaaS lock-in](https://stratafs.neullabs.com/blog/self-hosted-search-without-saas-lockin): Most 'AI search' tools want your data on their servers. We didn't want to send our files anywhere, so we built StrataFS. Here's the case for self-hosted semantic search.
- [SQLite as a hybrid search engine: FTS5, sqlite-vec, and embedded retrieval](https://stratafs.neullabs.com/blog/sqlite-fts5-vector-embedded-search): The 'embedded search server' used to be a contradiction. With SQLite FTS5 for BM25 and sqlite-vec for cosine similarity, both indexes live in one file — and one SQL query fuses them.
- [Hybrid search: combining BM25 and vector ranking in one SQL query](https://stratafs.neullabs.com/blog/hybrid-search-bm25-vector-explained): Hybrid search isn't a gimmick — it solves real failure modes of pure BM25 and pure vector retrieval. Here's the math, the SQL, and the tuning knobs that matter.
- [Building AI agents that understand your knowledge base](https://stratafs.neullabs.com/blog/mcp-server-for-knowledge-bases): The Model Context Protocol is the USB-C of agent integrations — but most MCP servers stop at the protocol. StrataFS goes further: hybrid retrieval, pre-shaped responses, per-source isolation, all local.
- [Semantic search for your codebase: an alternative to grep](https://stratafs.neullabs.com/blog/semantic-search-for-codebases): Grep is precise but literal. Semantic code search finds the file you mean, not the file you spelled. Here's why hybrid retrieval beats either alone — and how to run it locally with StrataFS.

## Comparisons

- [StrataFS vs. Elasticsearch — a detailed comparison](https://stratafs.neullabs.com/compare/elasticsearch): Both index files and serve search. Elasticsearch scales to billions of documents in a cluster; StrataFS runs as a single binary with SQLite on disk. Here's when each is right.
- [StrataFS vs. grep — semantic search where grep falls short](https://stratafs.neullabs.com/compare/grep): Grep and ripgrep are precision instruments for exact strings. StrataFS adds semantic understanding for the queries grep can't answer. Use both — they're complements, not competitors.
- [StrataFS vs. Pinecone — when to skip the managed vector DB](https://stratafs.neullabs.com/compare/pinecone): Pinecone is a managed vector database. StrataFS is a self-hosted semantic filesystem with vector search built in. Here's when each makes sense — and when Pinecone is solving a problem you don't have.