LlamaIndex + MCP Usage¶

The llama-index-tools-mcp package provides several tools for using MCP with LlamaIndex.

In [ ]:

%pip install llama-index-tools-mcp

%pip install llama-index-tools-mcp

Using Tools from an MCP Server¶

Using the get_tools_from_mcp_url or aget_tools_from_mcp_url function, you can get a list of FunctionTools from an MCP server.

In [ ]:

from llama_index.tools.mcp import (
    get_tools_from_mcp_url,
    aget_tools_from_mcp_url,
)

# async
tools = await aget_tools_from_mcp_url("http://127.0.0.1:8000/sse")

from llama_index.tools.mcp import ( get_tools_from_mcp_url, aget_tools_from_mcp_url, ) # async tools = await aget_tools_from_mcp_url("http://127.0.0.1:8000/sse")

By default, this will use our BasicMCPClient, which will run a command or connect to the URL and return the tools.

You can also pass in a custom ClientSession to use a different client.

You can also specify a list of allowed tools to filter the tools that are returned.

In [ ]:

from llama_index.tools.mcp import BasicMCPClient

client = BasicMCPClient("http://127.0.0.1:8000/sse")

tools = await aget_tools_from_mcp_url(
    "http://127.0.0.1:8000/sse",
    client=client,
    allowed_tools=["tool1", "tool2"],
)

from llama_index.tools.mcp import BasicMCPClient client = BasicMCPClient("http://127.0.0.1:8000/sse") tools = await aget_tools_from_mcp_url( "http://127.0.0.1:8000/sse", client=client, allowed_tools=["tool1", "tool2"], )

Converting a Workflow to an MCP App¶

If you have a custom Workflow, you can convert it to an MCP app using the workflow_as_mcp function.

For example, let's use the following workflow that will make a string loud:

In [ ]:

from llama_index.core.workflow import (
    Context,
    Workflow,
    Event,
    StartEvent,
    StopEvent,
    step,
)
from llama_index.tools.mcp.utils import workflow_as_mcp


class RunEvent(StartEvent):
    msg: str


class InfoEvent(Event):
    msg: str


class LoudWorkflow(Workflow):
    """Useful for converting strings to uppercase and making them louder."""

    @step
    def step_one(self, ctx: Context, ev: RunEvent) -> StopEvent:
        ctx.write_event_to_stream(InfoEvent(msg="Hello, world!"))

        return StopEvent(result=ev.msg.upper() + "!")


workflow = LoudWorkflow()

mcp = workflow_as_mcp(workflow)

from llama_index.core.workflow import ( Context, Workflow, Event, StartEvent, StopEvent, step, ) from llama_index.tools.mcp.utils import workflow_as_mcp class RunEvent(StartEvent): msg: str class InfoEvent(Event): msg: str class LoudWorkflow(Workflow): """Useful for converting strings to uppercase and making them louder.""" @step def step_one(self, ctx: Context, ev: RunEvent) -> StopEvent: ctx.write_event_to_stream(InfoEvent(msg="Hello, world!")) return StopEvent(result=ev.msg.upper() + "!") workflow = LoudWorkflow() mcp = workflow_as_mcp(workflow)

This code will automatically generate a FastMCP server that will

• Use the workflow class name as the tool name
• Use our custom RunEvent as the typed inputs to the tool
• Automatically use the SSE stream for streaming json dumps of the workflow event stream

If this code was in a script called script.py, you could launch the MCP server with:

mcp dev script.py

Or the other commands documented in the MCP CLI README.

Note that to launch from the CLI, you may need to install the MCP CLI:

pip install "mcp[cli]"

You can further customize the FastMCP server by passing in additional arguments to the workflow_as_mcp function:

• workflow_name: The name of the workflow. Defaults to the class name.
• workflow_description: The description of the workflow. Defaults to the class docstring.
• start_event_model: The event model to use for the start event. You can either use a custom StartEvent class in your workflow or pass in your own pydantic model here to define the inputs to the workflow.
• **fastmcp_init_kwargs: Any extra arguments to pass to the FastMCP() server constructor.