Latest insights, tutorials, and updates from the Model Context Protocol team
kratos is a Go-based microservices framework for cloud-native AI applications, functioning as an MCP Server. It provides a robust foundation for building scalable services in Go, ideal for developers targeting modern distributed systems and integrating with the Model Context Protocol.
agents is an MCP Client providing an agentic plugin marketplace with 83 plugins, 191 agents, 155 skills, and 102 commands. It targets developers building production-ready workflows across Claude Code, OpenAI Codex CLI, Cursor, OpenCode, and Gemini CLI, all from a single Markdown source.
headroom is an MCP Server that dramatically compresses tool outputs, logs, files, and RAG chunks before they reach the LLM, achieving 60-95% token reduction. Developers working with LLMs who need to optimize context window usage will find its MCP capabilities useful.
ai-berkshire is an MCP Client that provides a value investing research framework, integrating methodologies from four masters via AI agents. It's designed for Claude Code and Codex users seeking to automate and deepen their investment analysis.
private-gpt is a Python-based MCP Client providing a complete API layer for private AI applications built on local models. It enables RAG, skills, and tools, allowing developers to build production AI products without cloud API dependencies, all while integrating with OpenAI-compatible inference servers.
codebase-memory-mcp is a high-performance C-language MCP Server designed for rapid code intelligence. It indexes codebases into a persistent knowledge graph, offering sub-millisecond queries and drastically reduced token usage for AI agents. Developers needing fast, dependency-free code analysis will find its MCP integration particularly useful.