What it does
officegrep extracts text from Word, PowerPoint, Excel and PDF files and stores it in a single SQLite index, which agents query over MCP. Each search returns the file path, a readable locator such as a page, slide or sheet cell, and a highlighted snippet. The server exposes search_docs, list_docs and index_folder tools. Indexing is incremental, and the SQLite FTS5 trigram tokenizer supports substring matching for both CJK and Latin text. Queries of one or two characters fall back to a slower scan.
It suits cases where an agent needs to find where a phrase appears in office or PDF documents that plain grep cannot read. It runs offline and needs no embedding service or vector store. The publisher's benchmark on a corpus of 43 files (2,260 text blocks) reports a first index time of 2.16 s and a 1.9 MiB database.
The index is stored at ~/.officegrep/index.db and can be moved with a flag or environment variable. If the index is empty when the MCP server starts, it scans Documents, Downloads and Desktop in the background, with a 25 MB per-file cap. That scan can be disabled or pointed at other folders. WAL mode lets a watch process and the MCP server share one database. The server works with Claude Desktop, Cursor, Codex and other MCP hosts.
Limitations and requirements
Requires the Python packages lxml and openpyxl; PDF support needs the optional pypdf. The .mcpb bundle needs uv. No OCR, so scanned PDFs yield no text. Search is literal, not semantic. One- and two-character queries scan the table and become slow past a few hundred thousand blocks. Only body text is indexed; comments, tracked changes, headers and footers are not.
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
officegrep
Full-text search inside Word, PPT, Excel and PDF files, with precise locators.
v0.1.3As shown on registry.modelcontextprotocol.io when we last read it