Show HN: Ocrisp, One-Click RAG Implementation, Simple and Portable (github.com)

Ocrisp is a lightweight, single-executable tool that implements a one-click RAG (retrieval-augmented generation) workflow with GUI, CLI and an MCP (client) mode. The basic flow is intentionally simple: drop PDFs into a local data folder, run the embedding step, then connect LLM clients that speak MCP to query the embedded documents. The project emphasizes portability and ease-of-use—everything lives in one binary so you can run it locally without complex orchestration. Technically, Ocrisp exposes an MCP server so frontends like Claude Desktop can retrieve vectors and perform retrieval over your PDFs; on Windows you register it by editing %APPDATA%\Claude\claude_desktop_config.json and adding an mcpServers entry that points to your ocrisp.exe with args ["mcp"]. It currently targets file-based embedding + vector storage and plans to support other vector databases (LanceDB mentioned). Significance: it lowers the barrier for quick, private RAG setups and rapid prototyping, especially for users who want an offline or self-contained solution. Caveat: it’s an early, single-use-case project with limited features today, so expect improvements and integrations to come.