The RLM MCP Server is a large-scale context processing tool based on the Recursive Language Model pattern. It allows Claude Code to process text of over 10 million tokens through external variables, avoiding directly inputting massive content into the prompt. Through the processes of loading, chunking, sub-querying, and aggregation, it supports automatic analysis and programmatic execution, and can connect to the Claude API or the local Ollama for free inference.

An MCP server based on the recursive language model pattern, which processes ultra - large - scale contexts (over 10 million tokens) through chunking, sub - queries, and local inference, supporting automatic analysis, code execution, and security filtering.