ip2a/mcpstore

What it solves

mcpstore provides a centralized management platform for Model Context Protocol (MCP) services. It eliminates the friction of manually configuring and running individual MCP servers by offering a unified way to handle their configuration, execution, lifecycle management, and integration into various AI agent frameworks.

How it works

Built with Rust, the project provides a core management layer that can be accessed via several interfaces:

• SDKs: Python and Rust libraries allow developers to programmatically manage services and integrate tools into their applications.
• CLI: A command-line tool for managing services and exposing them as HTTP or stdio MCP servers.
• App/Web: A graphical interface for visualizing sessions, managing skills, and organizing agents.
• Framework Adapters: It includes built-in adapters that translate MCP tools into formats compatible with popular frameworks like LangChain, LangGraph, OpenAI, AutoGen, CrewAI, LlamaIndex, and Semantic Kernel.
• Data Sharing: It supports KV backends like Redis to share service and tool data across multiple processes or instances.

Who it’s for

Developers building AI agents who need a scalable way to manage multiple MCP servers and easily plug those capabilities into different LLM orchestration frameworks.

Highlights

• Multi-Framework Support: Seamlessly converts MCP tools for use in LangChain, OpenAI, and other major agent libraries.
• Agent Scoping: Allows creating independent scopes for different agents, ensuring each agent has its own specific set of tools and services.
• Lifecycle Management: Provides comprehensive API controls to add, update, restart, and disconnect services.
• Session Transfer: Enables the transfer of sessions between different agents or workspaces to maintain context continuity.