Kensaku MCP
Kensaku searches names, text inside documents and screenshots locally and opens the same index to agents over MCP.
- 1,742 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The agent receives contents of indexed files, including personal documents.
- Installers are not code-signed.
Install
- Works in
Manual install
{
"mcpServers": {
"kensaku": { "command": "kensaku-cli", "args": ["mcp"] }
}
}Snippet from the Connect AI agents (MCP) section. `kensaku-cli` ships with every release, the app is downloaded from the releases page.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The agent receives contents of indexed files, including personal documents.
- Installers are not code-signed.
This is third-party code. Review the repository files before installing.
What it does
Kensaku is an app and CLI for local file search. One index covers names, text inside PDF, DOCX, EPUB and code, text recognized in screenshots (OCR) and search by meaning. The `kensaku-cli mcp` command starts an MCP server so an agent searches the same index. Responses are size-limited and file content arrives as data, not instructions. `kensaku-cli serve` listens on loopback only and requires a token from the config directory. Queries support filters such as `type:pdf`, `ext:md`, `modified:7d`, `size:>10mb`, `path:` and `re:` regular expressions. The interface is in English and Simplified Chinese, with light and dark themes.
Who it is for. People who lose files among thousands of documents and screenshots and want Claude Code or Cursor to find them without uploading folders to a cloud.
Good fit when
- You need a file whose meaning you remember but not its name.
- An error line survives only in a screenshot with no useful name.
- An agent looks up notes and PDFs across project folders without reading the whole disk.
- Searching Chinese, Japanese and Korean documents alongside English ones.
Not a fit when
- Files live on network drives that must not be indexed.
- You need shared cloud search for a team.
- An index with document contents cannot be stored on the work machine.
Example request
Find the PDF about database indexing I saved last week and show its path.MCP
- Transport
- stdio, http
- Authentication
- not required
Limitations
Builds are not code-signed yet, so macOS and Windows show a warning on first launch. Screenshot OCR and search by meaning need separate packs, optional at first run. The index stores fragments of your file contents on disk, which matters on shared computers. First indexing takes time: on the reference Linux machine it ran 184 seconds for 101 thousand mixed files.
How to disable. Remove the `kensaku` entry from the client MCP configuration and stop `kensaku-cli serve` if it runs. Delete the index directory in the OS config location separately.
FAQ
How do I connect Kensaku to an agent?
Add a `kensaku` server with the `kensaku-cli` command and the `mcp` argument to the MCP configuration. For HTTP clients run `kensaku-cli serve`.
Does any data leave the machine?
No. The README states there is no account, telemetry or cloud search, and the index sits in the OS config directory.
Can an agent modify files?
Access is read-only by default.
Is there a console mode?
Yes. `kensaku-cli search "docker error"` searches from a terminal and `--json` gives machine output.
Which systems have builds?
Windows (Intel/AMD and ARM), macOS (universal image) and Linux (AppImage, deb, rpm). Every release includes a `SHA256SUMS` file for verification.
Discussion
No questions or reviews yet.
Related
MCP serversA fast file search MCP server: Everything SDK on Windows, Spotlight on macOS and locate on Linux behind one search tool.
An MCP server for Kaiten: cards, comments, spaces and boards with control over response verbosity
An MCP server that lets an agent work on your computer: terminal and long-running processes, file search and edits, Excel, PDF and DOCX
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
Kaiten MCP with documents
kaiten-mcp
A 66-tool Kaiten MCP server, the only one in this batch supporting Kaiten documents and their inline comment threads
An MCP server and Mac app that keeps one local index of repos, agent sessions and tickets, with semantic search and work threads
README in short
The app runs in the tray and keeps the index warm. The README measures name search across 100,000 files in milliseconds on a warm cache.