MCP — Model Context Protocol

Expose any MCP server's tools to your agent — the whole MCP ecosystem for free. Requires pip install "toolloop[mcp]".

Connecting

from toolloop.mcp import McpServerConfig, mcp_tools

config = McpServerConfig(command="uvx", args=["mcp-server-fetch"])
# or streamable HTTP:
# config = McpServerConfig(url="https://example.com/mcp", headers={...})

async with mcp_tools(config) as tools:  # also accepts a list of configs
    agent = Agent(provider, tools=tools)  # tools are alive only inside the with
    await agent.run("fetch the toolloop README")

The context manager opens the transports (stdio subprocess or HTTP), performs the MCP handshake, discovers the tools and yields ToolDefinitions. Sessions close when the block exits — run your agent inside the async with.

Multiple servers

tools = []
async with mcp_tools(
    [
        McpServerConfig(command="uvx", args=["mcp-server-fetch"]),
        McpServerConfig(url="https://example.com/mcp", prefix="corp_"),
    ]
) as tools:
    ...

The optional prefix avoids name collisions across servers.

Pass-through arguments

Arguments flow to the server untouched: the server's inputSchema is rendered verbatim into the system prompt, and validation stays server-side (per the MCP specification). Errors reported by the server (isError) become repair observations for the model.

Advanced: from an existing session

from toolloop.mcp import mcp_tools_from_session

tools = await mcp_tools_from_session(session, dangerous=False, prefix="")

Use it when you manage the ClientSession yourself (or in tests).

Example

A fully offline example — it spawns its own MCP server as a subprocess and connects through the real stdio transport — lives in examples/06_mcp_tools.py.