mcpstore

McpStore is an MCP service orchestration layer designed to help developers manage multiple MCP services in a unified store and expose them as tools for AI frameworks like LangChain. It provides a store abstraction that can host diverse MCP configurations, supports grouping by agent, and offers convenient methods to list, add, update, or delete services and tools. The framework integrates with LangChain and other tooling ecosystems, enabling you to convert MCP services into reusable tools or agents that can be invoked programmatically. Start by initializing a store, then add an MCP service configuration (which can be provided in multiple mainstream MCP formats) and query or render the corresponding tools for LangChain or other frameworks.

Using the Python API, you typically create a store with MCPStore.setup_store(), then add services via store.for_store().add_service(...) and convert them into LangChain tools with store.for_store().for_langchain().list_tools(). This lets you quickly assemble a set of MCP-backed tools that can be invoked by a LangChain agent. You can also organize services per agent to isolate contexts, and you can perform common operations such as finding, updating, restarting, or deleting services, as well as checking health and listing available tools. The project emphasizes compatibility with multiple MCP formats so you can reuse existing configurations with minimal friction.

Prerequisites:

• Python 3.8 or newer
• Internet access to download packages

Install the MCPStore package from PyPI:

pip install mcpstore

(Optional) If you plan to share state across processes or machines, you can install Redis support and configure an external Redis database:

pip install mcpstore[redis]

Basic usage to run the API server (via CLI) after installation:

mcpstore run api

If you prefer programmatic startup, you can initialize the store in Python:

from mcpstore import MCPStore
store = MCPStore.setup_store()
store.start_api_server(host="0.0.0.0", port=18200)

For more configuration options, see the project's documentation and API references.

Environment and configuration tips:

• You can pass an external Redis backend for cross-process state by configuring external_db in setup_store, e.g., external_db={"cache": {"type": "redis", "url": "redis://localhost:6379/0", "namespace": "demo_namespace"}}
• Tools are dynamically generated from MCP service configurations; you can mix formats and still get a usable toolset if they conform to MCP patterns understood by mcpstore.
• When deploying, ensure your runtime environment has network access to any MCP endpoints and that API ports (default 18200 in examples) are open.
• If you encounter startup issues, check that Python path resolution is correct and that the mcpstore package version is compatible with your Python version.