Save it once. Find it by meaning.

Recall is a quiet home for notes, snippets, and pages worth keeping. Search works on what you meant — not just the words you used. And your AI assistant can search it too.

Open Recall View on GitHub

Ask Recall about Recall.

live demo

Six notes saved about Recall — the real thing searches everything you save.

Or try one of these:

Model Context Protocol

Give your AI a memory.

Recall speaks MCP — the open protocol AI assistants use to reach outside tools. Connect it once, and your assistant can search your notes mid-conversation, follow threads between them, and save new things as you work. No copy-paste, no exports.

search_notesyour assistant finds what you know
add_noteit saves things for you
get_relatedit follows threads between notes

Works with Claude Desktop, Cursor, and any MCP-capable client.

Which caching approach did I settle on for the API?

search_notes("API caching approach")mcp

A Redis read-through cache that fails open — if the cache is down, requests fall through to Postgres and get slower, never fail. You noted warm searches came back about 200× faster.

Good — note that we’re keeping the sidecar setup.

add_note("Keeping Redis as an in-task sidecar…")mcp

Saved to Recall.

Capture without friction

Write a note, paste a URL and Recall extracts the readable article, or bulk-import a folder of Markdown.

Search that gets the idea

Hybrid retrieval pairs semantic understanding with exact keyword matching — half-remembered concepts and precise phrases both land.

Private by default

Under the hood

hybrid retrieval — vectors + full-text
reciprocal rank fusion
Postgres + pgvector
FastAPI on AWS
MCP over streamable HTTP
open source

Read the code on GitHub →

Start remembering.

Your next good idea shouldn’t disappear into a doc you’ll never open again.

Open Recall