Build effective agents with Model Context Protocol using simple, composable patterns.
mcp-agent is a simple, composable framework to build agents using Model Context Protocol.
We believe that MCP is all you need to build agents. In practice, that means:
- Agents connect to the outside world via MCP. Tools, resources, and prompts are exposed via MCP servers, leading to simple agent design.
- Agents expose themselves over MCP. Agents themselves can be deployed as MCP servers, so any MCP client can invoke your agent, and agents can even invoke other agents over the same vendor-neutral protocol.
Inspired by Anthropic’s foundational Model Context Protocol, which standardizes how software interfaces with AI assistants, and the influential guide “Building Effective Agents,” which outlines practical patterns for agent development.
MCP Agent Cloud
When you're ready to run in production, MCP Agent Cloud lets you deploy agents as remote MCP servers(SSE/streaming HTTP) and execute tools as durable, long-running, fault-tolerant (Temporal) workflows.
Quick Example:
Follow the commands in the terminal view for a quick and hassle free setup.
main.py
OUTPUTTERMINALPORTS
uvx mcp-agent init
# initialize mcp-agent project
uv init
# initialize uv project
uv add "mcp-agent[openai]"
# install mcp-agent pkg
uv run main.py
# run basic example locally
uvx mcp-agent deploy
# deploy example as an MCP server