Kensaku logo

Kensaku

Visit

Kensaku is a local, open-source desktop search launcher that indexes files, documents, screenshots, and code, with read-only MCP access for AI agents.

Share:
View alternatives

Kensaku

Kensaku is a MIT-licensed desktop launcher and local search index for Windows, macOS, and Linux. It is useful when a filename is not enough: its maintainers document filename, full-text, fuzzy, and meaning-based search across documents, code, logs, images, and screenshots. The project also provides a CLI and a read-only MCP interface so a compatible agent can search the same local index.

What it does

  • Indexes selected folders incrementally and searches names, extracted text, OCR text, and semantic matches.
  • Supports common office documents, PDFs, EPUB, source code, logs, and images. The documentation lists PP-OCRv6 as the default OCR engine, with optional Indra and Tesseract choices.
  • Provides filters for type, path, time, size, and content, plus a kensaku-cli command for scripts.
  • Exposes kensaku-cli mcp for local MCP clients. The documented agent tools are read-only by default; its HTTP mode listens on loopback and uses a bearer token.

A practical workflow

Install the official release, begin with one disposable project or document folder, and set exclusions before the first scan. Test three searches: a remembered phrase, text from a screenshot, and a natural-language description. If results look appropriate, connect an MCP client and ask it to search for files rather than granting it broad filesystem access. Confirm returned paths and source text yourself before acting on an agent’s answer.

Privacy and limits

Kensaku says it has no account, telemetry, or cloud search, but the local index contains filenames and extracted text. Treat that index as sensitive and do not sync it to a shared location without review. The project says current builds are not code-signed, so Windows and macOS may display an initial warning. OCR packs and semantic features can require additional downloads, and large folders still need local CPU, memory, and disk time to index.

Alternatives and next step

DigUp is another local semantic search option, while PaperPilot focuses on bounded document operations rather than whole-library retrieval. Download Kensaku from its official releases, verify the supplied checksums, and evaluate it on a non-sensitive folder first.

Comments

No comments yet. Be the first to comment!