Provides a lightweight, local-first server for semantic search, file chunking, and embeddings with a web UI and API.
About
PocketMCP is a lightweight, local-first MCP (Model Context Protocol) server that automatically watches folders, chunks and embeds files locally using Transformers.js with MiniLM, stores vectors in SQLite + sqlite-vec, and exposes semantic search capabilities to VS Code and Cursor. Designed for small machines (I'm running on an Intel N100 with 16GB RAM) with zero external dependencies after initial model download.
Key Features
- Automated file ingestion and processing for multiple formats (PDF, DOCX, Markdown, text).
- Semantic search for content by meaning, not just keywords.
- Built-in web interface and API for testing, validation, and manual document management.
- Native MCP integration for VS Code and Cursor, supporting both stdio and HTTP transports.
- 0 GitHub stars
- Local-first operation with offline capabilities and efficient SQLite vector storage.
Use Cases
- Providing a self-hosted, resource-efficient semantic search backend for custom web or desktop applications.
- Enhancing AI coding assistants (VS Code, Cursor) with local-first semantic search capabilities.
- Building personal or team-specific knowledge bases from various document types for rapid content retrieval.
Advertisement