semanticsearch — Claude Code plugin: search and install
A local MCP server indexes your documents and answers natural-language searches. It works with PDF, DOCX, Markdown, TXT, PPTX, and CSV files.
/plugin marketplace add anthropics/claude-plugins-community
/plugin install semanticsearch@claude-communityWhat semanticsearch does
semanticsearch gives Claude Code access to a local document index. It extracts text, splits it into smaller passages, creates embeddings, and stores the results in a LanceDB file database. The MCP server exposes index_folder, semantic_search, get_index_status, and reindex_file.
Queries can use vector similarity, keyword matching, or a hybrid of both. SHA-256 checks let incremental indexing leave unchanged files alone. Supported inputs include .txt, .md, .pdf, .docx, .pptx, and .csv; the embedding model is downloaded on first use and is about 120MB.
Install it from the Claude Code plugin marketplace, then let the client connect to the MCP server. Python 3.11+ is required, so an older Python runtime will stop the setup from working.
What it helps with
- index a folder of local documents
- search across several files at once
- find passages in PDF and DOCX files
- inspect the current index contents
- rebuild the index for one file
Example prompts for the agent
- Index
~/Documents/projectsand leave unchanged files alone. - Search the indexed documents for details about API rate limits using hybrid mode.
- Show the index status, then force a rebuild of
notes/report.pdf.
How to install semanticsearch
Plugin marketplace
/plugin marketplace add anthropics/claude-plugins-community/plugin install semanticsearch@claude-communityStep by step, with commands for each platform: Claude Code plugins: marketplaces, installing, and building your own
Questions about semanticsearch
How do I install semanticsearch in Claude Code?
Run /plugin marketplace add anthropics/claude-plugins-community, then run /plugin install semanticsearch@claude-community.
What does semanticsearch do?
It builds a local index from supported documents and retrieves relevant passages for natural-language questions. Retrieval can combine embedding similarity with keyword matches.
What does semanticsearch need to run?
The project requires Python 3.11+. Its embedding model downloads on the first run and is about 120MB; later searches run on the local machine.
