Chaty vs ref-tools-mcp
Chaty: Chaty is a private, 100% offline desktop application that runs open LLMs directly on your machine. It integrates various AI functionalities including a local coding agent, document knowledge base (RAG), deep research capabilities, and hands-free voice interaction, all with GPU acceleration.; ref-tools-mcp: Ref MCP is a server that enables AI coding tools and agents to access technical documentation efficiently. It optimizes token usage by filtering search results and extracting only the most relevant parts of documentation pages, helping models stay focused and reducing operational costs.
TL;DR
Secure and private software development and debugging using an AI coding agent that interacts with your local codebase.
- Context size is your main concern: page reads return about the most relevant 5k tokens, and repeat searches do not return the same results in a session.
- You want two simple tools, search and read-URL, in Claude Code, Cursor or another MCP client.
- You want to search private resources such as repos and PDFs alongside public docs.
- You cannot send queries to a vendor: it needs an API key and the hosted ref.tools service.
- You want the recommended HTTP server: this repository is the legacy stdio server.
- You need independent proof of the token savings: the README's numbers are the project's own example traces.
Side-by-Side Comparison
Features
Use Cases
Best For
FAQ
Deploy on DigitalOcean
Host your MCP servers, Docker containers, and AI agents with $200 free credit over 60 days.