Kensaku: Open-Source Local Search and File Launcher with OCR

Kensaku is a free, open-source local file launcher and search tool designed to find files across documents, code, and images entirely on a user's machine. It combines filename matching, full-text search, typo-tolerant fuzzy search, and semantic capabilities into a single application supporting Windows, macOS, and Linux.

Core Features and Capabilities

• Local Indexing: Operates entirely offline with no telemetry or mandatory cloud account, keeping indices within the operating system's configuration directory.
• Multi-Format Support: Reads standard documents (PDF, DOCX, XLSX, EPUB), 50+ programming languages and data formats, emails, and images using OCR (bundled PP-OCRv6, with optional Indra or Tesseract engines).
• Flexible Interfaces: Runs either as a tray launcher summoned via a global keystroke or as a traditional desktop window, featuring light and dark themes and space previews.
• AI Agent Integration: Connects with AI tools like Claude and Cursor through the Model Context Protocol (MCP) or loopback HTTP for read-only local index searching.

Performance and Security

Kensaku utilizes incremental indexing to process changed files without full-disk rescanning. Its memory footprint for metadata remains compact, using an in-memory layer of roughly 12.8 MB for 100,000 file entries. Privacy measures include skipping hidden files, honoring `.gitignore` and `.cursorignore` rules, and excluding sensitive credential patterns by default.

Installation and Verification

Releases are distributed with SHA256SUMS files to verify downloads. Because builds are currently unsigned, operating systems like Windows and macOS may display security prompts during the initial run, which users can bypass through system settings. Source code and documentation are available via the project's repository for users wishing to build from source or contribute.