---
id: mcp-use
version: "1.7.0"
license: MIT
license_treatment: permissive
maintenance: active
---

# mcp-use — Full Stack MCP framework for python, build MCP agents, clients, and servers.

License: permissive · Maintenance: active · Downloads: 420.5K/mo

## What it is and what it does

mcp-use is a Python framework that bridges LLMs and the Model Context Protocol (MCP), a standard for connecting AI models to external tools and data sources. It provides three main components: an MCP Agent that reasons across multiple steps and selects tools to solve problems, an MCP Client for direct programmatic access to MCP servers, and an MCP Server builder for creating and deploying custom servers.

The package is designed for developers building AI applications that need tool access. It works with langchain-supported LLMs that support tool calling and handles multiple MCP servers simultaneously, allowing agents to dynamically select the appropriate server for a given task. It includes features like tool restrictions for safety, HTTP support for remote servers, and custom agent building via langchain adapters.

Use it for:

- Build an AI agent that can call multiple external APIs or services by connecting them as MCP servers.
- Create a chatbot that dynamically selects and executes tools from a pool of MCP servers based on user requests.
- Develop a custom MCP server to expose internal tools or data sources to LLM-powered applications.
- Integrate file system, database, or network tools into an LLM workflow with built-in safety restrictions.
- Prototype AI agent logic using the interactive code builder before deploying to production.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

mcp-use provides a Python framework to connect LLMs to Model Context Protocol servers, enabling agents to discover and call tools, clients to access MCP servers directly, and developers to build and deploy MCP servers.

Yes, if you are building LLM-powered applications that need tool access and want a structured framework for MCP integration. The low install friction, active maintenance, permissive license, and support for multiple LLM providers make it a solid choice. The Alpha status and 11 dependencies warrant testing in a non-critical environment first, but the lack of known vulnerabilities and recent release cycle suggest reasonable stability for early adoption.

## Install

pip install mcp-use
uv add mcp-use
poetry add mcp-use

## Installing mcp-use

Before you install: Low install friction with a pure-wheel distribution. Actively maintained with a recent release. Requires 11 runtime dependencies including langchain, pydantic, and httpx, which are themselves stable and widely used.

License in practice: MIT license is permissive, allowing commercial and private use with minimal restrictions; you must include a copy of the license in distributions.

Quickstart:

pip install mcp-use

from mcp_use import MCPAgent

agent = MCPAgent(llm=None)
# Connect to MCP servers and use tools

Requires Python 3.11 or later. Requires a langchain-compatible LLM provider configured via environment or passed to the agent.

Verify before relying:

- Whether the 'six lines of code' claim for a first agent is accurate or marketing language.
- Specific performance characteristics or latency for multi-server agent scenarios.
- Production readiness status beyond the 'Alpha' development classifier.
- Whether tool restrictions (file system, network access) are enforced at runtime or advisory only.
- Which specific langchain-supported LLM providers are tested and recommended.

## Package facts

- License: MIT (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 420.5K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

model context protocol python, llm tool calling framework, mcp agent builder, connect llm to mcp server, ai agent with tool access, mcp client library, build mcp servers python, llm-integration, agent-framework, tool-calling

[View on SkillFed](https://skillfed.io/packages/mcp-use) · [View on PyPI](https://pypi.org/project/mcp-use/)