Your documents should be searchable without being surrendered.
Alcove is a local-first document search library. Install it, point it at your files, and search. No server, no sign-up, no data leaves your disk.
It indexes PDFs, EPUB, HTML, Markdown, CSV, JSON, DOCX, and plain text into a local vector store. Queries stay on your machine. Results come from your machine. There is no cloud.
How it works
ingest
Point at a directory. Alcove extracts text and chunks it.
index
Embeddings and metadata go into a local vector store.
query
Search via CLI, web UI, or REST API. All local.
Get started
$ pip install alcove-search
$ alcove seed-demo # download sample corpus + build index
$ alcove serve # open http://localhost:8000
Requires Python 3.10+. The seed demo indexes five public-domain texts (Alice in Wonderland, Frankenstein, the Federalist Papers, the Declaration of Independence, the Constitution) so you can search immediately.
Trust model
We do not want your data.
All documents stay on your disk. There is no hosted control plane.
No telemetry. ChromaDB’s upstream telemetry is also disabled.
No outbound network calls by default. You choose what enters your index.
Apache 2.0 licensed. Read the source.
Built for
Small archives
Personal collections that deserve real search.
Journalism
FOIA dumps without a SaaS contract.
Research
Papers and notes, indexed on your terms.
Heritage collections
Digitized records that belong to communities, not platforms.
Streaming corpora
Ingest as documents arrive. The index grows with you.
AI workflows
Local retrieval for agents that respect boundaries.
Agent-native
Alcove exposes a REST API out of the box. Designed for integration with AI agents and tool-use workflows : MCP endpoint coming.
MCP · planned