A local MCP server that manages LLM session memory using typed knowledge graphs. It automatically extracts tasks and decisions, prevents context overflow through checkpointing, and resumes sessions efficiently in ~250 tokens.
Tokenmizer is an MCP (Model Context Protocol) server that provides graph-structured session memory for large language models. It acts as a local OpenAI-compatible proxy that intelligently extracts and organizes tasks, decisions, and files into a typed knowledge graph, enabling LLMs to maintain persistent context across conversations without token waste.
Clone the repository from GitHub at ShwetaMishraai/tokenmizer. Install dependencies using Python package management. Configure it as an MCP server compatible with your AI platform. The tool works as a local proxy, so ensure your LLM client can route requests through it. Platform support includes ChatGPT, Claude, AI Studio, and other MCP-compatible services.
Monday.com MCP Server streamlines board management, item operations, and workflow automation for teams. I…
by NotionFlow
Sentry MCP Server provides comprehensive error tracking and performance monitoring, helping developers id…
by AnalyticsPro
Cloudflare MCP Server simplifies Cloudflare management by providing tools for DNS management, Workers dep…
by PricingBot