What it does
claude-find indexes the JSONL session files that Claude Code stores under ~/.claude/projects/. It extracts user and assistant messages, compact summaries and file paths from tool calls, and adds project, branch, file and date metadata to each chunk. Chunks are embedded with the qwen3-embedding model through Ollama, and searches combine semantic matching with keyword (FTS5) matching using Reciprocal Rank Fusion. The server returns raw conversation chunks rather than summaries.
It suits cases where Claude needs earlier reasoning, constraints, failed approaches or decisions from previous work. Tool call metadata is also searchable, for example by files touched or errors encountered. Existing sessions are covered without adding hooks.
Setup starts Ollama, downloads the embedding model, turns off Claude Code's 30-day session cleanup so sessions stay searchable, and registers the server with Claude Code. Indexing runs in the background at startup, so searches return increasingly complete results while it continues. After an upgrade, a reindex command rebuilds the index.
Limitations and requirements
Requires the Bun runtime, Ollama (the embedding model is downloaded on first use) and Claude Code. Setup changes Claude Code's setting that deletes sessions after 30 days.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
io.github.Cavinooo/claude-find
Pull Deep Memory from across your Claude Code Sessions — when you need it.
v0.2.5As shown on registry.modelcontextprotocol.io when we last read it