xsecuremcp / src /fastmcp /server /server.py
Jeremiah Lowin
add transport kwargs
9f94e73
Raw
History Blame
28.9 kB
"""FastMCP - A more ergonomic interface for MCP servers."""
import json
from collections.abc import AsyncIterator, Awaitable, Callable
from contextlib import (
AbstractAsyncContextManager,
AsyncExitStack,
asynccontextmanager,
)
from typing import TYPE_CHECKING, Any, Generic, Literal
import anyio
import httpx
import pydantic_core
import uvicorn
from fastapi import FastAPI
from mcp.server.lowlevel.helper_types import ReadResourceContents
from mcp.server.lowlevel.server import LifespanResultT
from mcp.server.lowlevel.server import Server as MCPServer
from mcp.server.session import ServerSession
from mcp.server.sse import SseServerTransport
from mcp.server.stdio import stdio_server
from mcp.types import (
AnyFunction,
EmbeddedResource,
GetPromptResult,
ImageContent,
TextContent,
)
from mcp.types import Prompt as MCPPrompt
from mcp.types import PromptArgument as MCPPromptArgument
from mcp.types import Resource as MCPResource
from mcp.types import ResourceTemplate as MCPResourceTemplate
from mcp.types import Tool as MCPTool
from pydantic.networks import AnyUrl
from starlette.applications import Starlette
from starlette.requests import Request
from starlette.routing import Mount, Route
import fastmcp
import fastmcp.settings
from fastmcp.exceptions import ResourceError
from fastmcp.prompts import Prompt, PromptManager
from fastmcp.prompts.prompt import Message, PromptResult
from fastmcp.resources import Resource, ResourceManager
from fastmcp.resources.template import ResourceTemplate
from fastmcp.tools import ToolManager
from fastmcp.tools.tool import Tool
from fastmcp.utilities.decorators import DecoratedFunction
from fastmcp.utilities.logging import configure_logging, get_logger
from fastmcp.utilities.types import Image
if TYPE_CHECKING:
from fastmcp.client import Client
from fastmcp.server.context import Context
from fastmcp.server.openapi import FastMCPOpenAPI
from fastmcp.server.proxy import FastMCPProxy
logger = get_logger(__name__)
@asynccontextmanager
async def default_lifespan(server: "FastMCP") -> AsyncIterator[Any]:
"""Default lifespan context manager that does nothing.
Args:
server: The server instance this lifespan is managing
Returns:
An empty context object
"""
yield {}
def lifespan_wrapper(
app: "FastMCP",
lifespan: Callable[["FastMCP"], AbstractAsyncContextManager[LifespanResultT]],
) -> Callable[
[MCPServer[LifespanResultT]], AbstractAsyncContextManager[LifespanResultT]
]:
@asynccontextmanager
async def wrap(s: MCPServer[LifespanResultT]) -> AsyncIterator[LifespanResultT]:
async with AsyncExitStack() as stack:
# enter main app's lifespan
context = await stack.enter_async_context(lifespan(app))
# Enter all mounted app lifespans
for prefix, mounted_app in app._mounted_apps.items():
mounted_context = mounted_app._mcp_server.lifespan(
mounted_app._mcp_server
)
await stack.enter_async_context(mounted_context)
logger.debug(f"Prepared lifespan for mounted app '{prefix}'")
yield context
return wrap
class FastMCP(Generic[LifespanResultT]):
def __init__(
self,
name: str | None = None,
instructions: str | None = None,
lifespan: (
Callable[["FastMCP"], AbstractAsyncContextManager[LifespanResultT]] | None
) = None,
tags: set[str] | None = None,
**settings: Any,
):
self.tags: set[str] = tags or set()
self.settings = fastmcp.settings.ServerSettings(**settings)
# Setup for mounted apps - must be initialized before _mcp_server
self._mounted_apps: dict[str, FastMCP] = {}
if lifespan is None:
lifespan = default_lifespan
self._mcp_server = MCPServer[LifespanResultT](
name=name or "FastMCP",
instructions=instructions,
lifespan=lifespan_wrapper(self, lifespan),
)
self._tool_manager = ToolManager(
duplicate_behavior=self.settings.on_duplicate_tools
)
self._resource_manager = ResourceManager(
duplicate_behavior=self.settings.on_duplicate_resources
)
self._prompt_manager = PromptManager(
duplicate_behavior=self.settings.on_duplicate_prompts
)
self.dependencies = self.settings.dependencies
# Set up MCP protocol handlers
self._setup_handlers()
# Configure logging
configure_logging(self.settings.log_level)
@property
def name(self) -> str:
return self._mcp_server.name
@property
def instructions(self) -> str | None:
return self._mcp_server.instructions
async def run_async(
self, transport: Literal["stdio", "sse"] | None = None, **transport_kwargs: Any
) -> None:
"""Run the FastMCP server asynchronously.
Args:
transport: Transport protocol to use ("stdio" or "sse")
"""
if transport is None:
transport = "stdio"
if transport not in ["stdio", "sse"]:
raise ValueError(f"Unknown transport: {transport}")
if transport == "stdio":
await self.run_stdio_async(**transport_kwargs)
else: # transport == "sse"
await self.run_sse_async(**transport_kwargs)
def run(
self, transport: Literal["stdio", "sse"] | None = None, **transport_kwargs: Any
) -> None:
"""Run the FastMCP server. Note this is a synchronous function.
Args:
transport: Transport protocol to use ("stdio" or "sse")
"""
logger.info(f'Starting server "{self.name}"...')
anyio.run(self.run_async, transport, **transport_kwargs)
def _setup_handlers(self) -> None:
"""Set up core MCP protocol handlers."""
self._mcp_server.list_tools()(self._mcp_list_tools)
self._mcp_server.call_tool()(self.call_tool)
self._mcp_server.list_resources()(self._mcp_list_resources)
self._mcp_server.read_resource()(self._mcp_read_resource)
self._mcp_server.list_prompts()(self._mcp_list_prompts)
self._mcp_server.get_prompt()(self._mcp_get_prompt)
self._mcp_server.list_resource_templates()(self._mcp_list_resource_templates)
def list_tools(self) -> list[Tool]:
return self._tool_manager.list_tools()
async def _mcp_list_tools(self) -> list[MCPTool]:
"""
List all available tools, in the format expected by the low-level MCP
server.
See `list_tools` for a more ergonomic way to list tools.
"""
tools = self.list_tools()
return [
MCPTool(
name=info.name,
description=info.description,
inputSchema=info.parameters,
)
for info in tools
]
def get_context(self) -> "Context[ServerSession, LifespanResultT]":
"""
Returns a Context object. Note that the context will only be valid
during a request; outside a request, most methods will error.
"""
try:
request_context = self._mcp_server.request_context
except LookupError:
request_context = None
from fastmcp.server.context import Context
return Context(request_context=request_context, fastmcp=self)
async def call_tool(
self, name: str, arguments: dict[str, Any]
) -> list[TextContent | ImageContent | EmbeddedResource]:
"""Call a tool by name with arguments."""
context = self.get_context()
result = await self._tool_manager.call_tool(name, arguments, context=context)
converted_result = _convert_to_content(result)
return converted_result
def list_resources(self) -> list[Resource]:
return self._resource_manager.list_resources()
async def _mcp_list_resources(self) -> list[MCPResource]:
"""
List all available resources, in the format expected by the low-level MCP
server.
See `list_resources` for a more ergonomic way to list resources.
"""
resources = self.list_resources()
return [
MCPResource(
uri=resource.uri,
name=resource.name or "",
description=resource.description,
mimeType=resource.mime_type,
)
for resource in resources
]
def list_resource_templates(self) -> list[ResourceTemplate]:
return self._resource_manager.list_templates()
async def _mcp_list_resource_templates(self) -> list[MCPResourceTemplate]:
"""
List all available resource templates, in the format expected by the low-level
MCP server.
See `list_resource_templates` for a more ergonomic way to list resource
templates.
"""
templates = self.list_resource_templates()
return [
MCPResourceTemplate(
uriTemplate=template.uri_template,
name=template.name,
description=template.description,
)
for template in templates
]
async def read_resource(self, uri: AnyUrl | str) -> str | bytes:
"""Read a resource by URI."""
resource = await self._resource_manager.get_resource(uri)
if not resource:
raise ResourceError(f"Unknown resource: {uri}")
return await resource.read()
async def _mcp_read_resource(self, uri: AnyUrl | str) -> list[ReadResourceContents]:
"""
Read a resource by URI, in the format expected by the low-level MCP
server.
See `read_resource` for a more ergonomic way to read resources.
"""
resource = await self._resource_manager.get_resource(uri)
if not resource:
raise ResourceError(f"Unknown resource: {uri}")
try:
content = await self.read_resource(uri)
return [ReadResourceContents(content=content, mime_type=resource.mime_type)]
except Exception as e:
logger.error(f"Error reading resource {uri}: {e}")
raise ResourceError(str(e))
def add_tool(
self,
fn: AnyFunction,
name: str | None = None,
description: str | None = None,
tags: set[str] | None = None,
) -> None:
"""Add a tool to the server.
The tool function can optionally request a Context object by adding a parameter
with the Context type annotation. See the @tool decorator for examples.
Args:
fn: The function to register as a tool
name: Optional name for the tool (defaults to function name)
description: Optional description of what the tool does
tags: Optional set of tags for categorizing the tool
"""
self._tool_manager.add_tool_from_fn(
fn, name=name, description=description, tags=tags
)
def tool(
self,
name: str | None = None,
description: str | None = None,
tags: set[str] | None = None,
) -> Callable[[AnyFunction], AnyFunction]:
"""Decorator to register a tool.
Tools can optionally request a Context object by adding a parameter with the
Context type annotation. The context provides access to MCP capabilities like
logging, progress reporting, and resource access.
Args:
name: Optional name for the tool (defaults to function name)
description: Optional description of what the tool does
tags: Optional set of tags for categorizing the tool
Example:
@server.tool()
def my_tool(x: int) -> str:
return str(x)
@server.tool()
def tool_with_context(x: int, ctx: Context) -> str:
ctx.info(f"Processing {x}")
return str(x)
@server.tool()
async def async_tool(x: int, context: Context) -> str:
await context.report_progress(50, 100)
return str(x)
"""
# Check if user passed function directly instead of calling decorator
if callable(name):
raise TypeError(
"The @tool decorator was used incorrectly. "
"Did you forget to call it? Use @tool() instead of @tool"
)
def decorator(fn: AnyFunction) -> AnyFunction:
self.add_tool(fn, name=name, description=description, tags=tags)
return DecoratedFunction(fn)
return decorator
def add_resource(self, resource: Resource) -> None:
"""Add a resource to the server.
Args:
resource: A Resource instance to add
"""
self._resource_manager.add_resource(resource)
def add_resource_fn(
self,
fn: AnyFunction,
uri: str,
name: str | None = None,
description: str | None = None,
mime_type: str | None = None,
tags: set[str] | None = None,
) -> None:
"""Add a resource or template to the server from a function.
If the URI contains parameters (e.g. "resource://{param}") or the function
has parameters, it will be registered as a template resource.
Args:
fn: The function to register as a resource
uri: The URI for the resource
name: Optional name for the resource
description: Optional description of the resource
mime_type: Optional MIME type for the resource
tags: Optional set of tags for categorizing the resource
"""
self._resource_manager.add_resource_or_template_from_fn(
fn=fn,
uri=uri,
name=name,
description=description,
mime_type=mime_type,
tags=tags,
)
def resource(
self,
uri: str,
*,
name: str | None = None,
description: str | None = None,
mime_type: str | None = None,
tags: set[str] | None = None,
) -> Callable[[AnyFunction], AnyFunction]:
"""Decorator to register a function as a resource.
The function will be called when the resource is read to generate its content.
The function can return:
- str for text content
- bytes for binary content
- other types will be converted to JSON
If the URI contains parameters (e.g. "resource://{param}") or the function
has parameters, it will be registered as a template resource.
Args:
uri: URI for the resource (e.g. "resource://my-resource" or "resource://{param}")
name: Optional name for the resource
description: Optional description of the resource
mime_type: Optional MIME type for the resource
tags: Optional set of tags for categorizing the resource
Example:
@server.resource("resource://my-resource")
def get_data() -> str:
return "Hello, world!"
@server.resource("resource://my-resource")
async get_data() -> str:
data = await fetch_data()
return f"Hello, world! {data}"
@server.resource("resource://{city}/weather")
def get_weather(city: str) -> str:
return f"Weather for {city}"
@server.resource("resource://{city}/weather")
async def get_weather(city: str) -> str:
data = await fetch_weather(city)
return f"Weather for {city}: {data}"
"""
# Check if user passed function directly instead of calling decorator
if callable(uri):
raise TypeError(
"The @resource decorator was used incorrectly. "
"Did you forget to call it? Use @resource('uri') instead of @resource"
)
def decorator(fn: AnyFunction) -> AnyFunction:
self._resource_manager.add_resource_or_template_from_fn(
fn=fn,
uri=uri,
name=name,
description=description,
mime_type=mime_type,
tags=tags,
)
return DecoratedFunction(fn)
return decorator
def add_prompt(
self,
fn: Callable[..., PromptResult | Awaitable[PromptResult]],
name: str | None = None,
description: str | None = None,
tags: set[str] | None = None,
) -> None:
"""Add a prompt to the server.
Args:
prompt: A Prompt instance to add
"""
self._prompt_manager.add_prompt_from_fn(
fn=fn,
name=name,
description=description,
tags=tags,
)
def prompt(
self,
name: str | None = None,
description: str | None = None,
tags: set[str] | None = None,
) -> Callable[[AnyFunction], AnyFunction]:
"""Decorator to register a prompt.
Args:
name: Optional name for the prompt (defaults to function name)
description: Optional description of what the prompt does
tags: Optional set of tags for categorizing the prompt
Example:
@server.prompt()
def analyze_table(table_name: str) -> list[Message]:
schema = read_table_schema(table_name)
return [
{
"role": "user",
"content": f"Analyze this schema:\n{schema}"
}
]
@server.prompt()
async def analyze_file(path: str) -> list[Message]:
content = await read_file(path)
return [
{
"role": "user",
"content": {
"type": "resource",
"resource": {
"uri": f"file://{path}",
"text": content
}
}
}
]
"""
# Check if user passed function directly instead of calling decorator
if callable(name):
raise TypeError(
"The @prompt decorator was used incorrectly. "
"Did you forget to call it? Use @prompt() instead of @prompt"
)
def decorator(func: AnyFunction) -> AnyFunction:
self.add_prompt(func, name=name, description=description, tags=tags)
return DecoratedFunction(func)
return decorator
async def run_stdio_async(self) -> None:
"""Run the server using stdio transport."""
async with stdio_server() as (read_stream, write_stream):
await self._mcp_server.run(
read_stream,
write_stream,
self._mcp_server.create_initialization_options(),
)
async def run_sse_async(
self,
host: str | None = None,
port: int | None = None,
log_level: str | None = None,
) -> None:
"""Run the server using SSE transport."""
starlette_app = self.sse_app()
config = uvicorn.Config(
starlette_app,
host=host or self.settings.host,
port=port or self.settings.port,
log_level=log_level or self.settings.log_level.lower(),
)
server = uvicorn.Server(config)
await server.serve()
def sse_app(self) -> Starlette:
"""Return an instance of the SSE server app."""
sse = SseServerTransport(self.settings.message_path)
async def handle_sse(request: Request) -> None:
async with sse.connect_sse(
request.scope,
request.receive,
request._send, # type: ignore[reportPrivateUsage]
) as streams:
await self._mcp_server.run(
streams[0],
streams[1],
self._mcp_server.create_initialization_options(),
)
return Starlette(
debug=self.settings.debug,
routes=[
Route(self.settings.sse_path, endpoint=handle_sse),
Mount(self.settings.message_path, app=sse.handle_post_message),
],
)
def list_prompts(self) -> list[Prompt]:
"""
List all available prompts.
"""
return self._prompt_manager.list_prompts()
async def _mcp_list_prompts(self) -> list[MCPPrompt]:
"""
List all available prompts, in the format expected by the low-level MCP
server.
See `list_prompts` for a more ergonomic way to list prompts.
"""
prompts = self.list_prompts()
return [
MCPPrompt(
name=prompt.name,
description=prompt.description,
arguments=[
MCPPromptArgument(
name=arg.name,
description=arg.description,
required=arg.required,
)
for arg in (prompt.arguments or [])
],
)
for prompt in prompts
]
async def get_prompt(
self, name: str, arguments: dict[str, Any] | None = None
) -> list[Message]:
"""Get a prompt by name with arguments."""
return await self._prompt_manager.render_prompt(name, arguments)
async def _mcp_get_prompt(
self, name: str, arguments: dict[str, Any] | None = None
) -> GetPromptResult:
"""
Get a prompt by name with arguments, in the format expected by the low-level
MCP server.
See `get_prompt` for a more ergonomic way to get prompts.
"""
try:
messages = await self.get_prompt(name, arguments)
return GetPromptResult(messages=pydantic_core.to_jsonable_python(messages))
except Exception as e:
logger.error(f"Error getting prompt {name}: {e}")
raise ValueError(str(e))
def mount(
self,
prefix: str,
app: "FastMCP",
tool_separator: str | None = None,
resource_separator: str | None = None,
prompt_separator: str | None = None,
) -> None:
"""Mount another FastMCP application with a given prefix.
When an application is mounted:
- The tools are imported with prefixed names using the tool_separator
Example: If app has a tool named "get_weather", it will be available as "weatherget_weather"
- The resources are imported with prefixed URIs using the resource_separator
Example: If app has a resource with URI "weather://forecast", it will be available as "weather+weather://forecast"
- The templates are imported with prefixed URI templates using the resource_separator
Example: If app has a template with URI "weather://location/{id}", it will be available as "weather+weather://location/{id}"
- The prompts are imported with prefixed names using the prompt_separator
Example: If app has a prompt named "weather_prompt", it will be available as "weather_weather_prompt"
- The mounted app's lifespan will be executed when the parent app's lifespan runs,
ensuring that any setup needed by the mounted app is performed
Args:
prefix: The prefix to use for the mounted application
app: The FastMCP application to mount
tool_separator: Separator for tool names (defaults to "_")
resource_separator: Separator for resource URIs (defaults to "+")
prompt_separator: Separator for prompt names (defaults to "_")
"""
if tool_separator is None:
tool_separator = "_"
if resource_separator is None:
resource_separator = "+"
if prompt_separator is None:
prompt_separator = "_"
# Mount the app in the list of mounted apps
self._mounted_apps[prefix] = app
# Import tools from the mounted app
tool_prefix = f"{prefix}{tool_separator}"
self._tool_manager.import_tools(app._tool_manager, tool_prefix)
# Import resources and templates from the mounted app
resource_prefix = f"{prefix}{resource_separator}"
self._resource_manager.import_resources(app._resource_manager, resource_prefix)
self._resource_manager.import_templates(app._resource_manager, resource_prefix)
# Import prompts from the mounted app
prompt_prefix = f"{prefix}{prompt_separator}"
self._prompt_manager.import_prompts(app._prompt_manager, prompt_prefix)
logger.info(f"Mounted app with prefix '{prefix}'")
logger.debug(f"Imported tools with prefix '{tool_prefix}'")
logger.debug(f"Imported resources with prefix '{resource_prefix}'")
logger.debug(f"Imported templates with prefix '{resource_prefix}'")
logger.debug(f"Imported prompts with prefix '{prompt_prefix}'")
@classmethod
async def as_proxy(
cls, client: "Client | FastMCP", **settings: Any
) -> "FastMCPProxy":
"""
Create a FastMCP proxy server from a client.
This method creates a new FastMCP server instance that proxies requests to the provided client.
It discovers the client's tools, resources, prompts, and templates, and creates corresponding
components in the server that forward requests to the client.
Args:
client: The client to proxy requests to
**settings: Additional settings for the FastMCP server
Returns:
A FastMCP server that proxies requests to the client
"""
from fastmcp.client import Client
from .proxy import FastMCPProxy
if isinstance(client, Client):
return await FastMCPProxy.from_client(client=client, **settings)
elif isinstance(client, FastMCP):
return await FastMCPProxy.from_server(server=client, **settings)
else:
raise ValueError(f"Unknown client type: {type(client)}")
@classmethod
def from_openapi(
cls, openapi_spec: dict[str, Any], client: httpx.AsyncClient, **settings: Any
) -> "FastMCPOpenAPI":
"""
Create a FastMCP server from an OpenAPI specification.
"""
from .openapi import FastMCPOpenAPI
return FastMCPOpenAPI(openapi_spec=openapi_spec, client=client, **settings)
@classmethod
def from_fastapi(
cls, app: FastAPI, name: str | None = None, **settings: Any
) -> "FastMCPOpenAPI":
"""
Create a FastMCP server from a FastAPI application.
"""
from .openapi import FastMCPOpenAPI
client = httpx.AsyncClient(
transport=httpx.ASGITransport(app=app), base_url="http://fastapi"
)
name = name or app.title
return FastMCPOpenAPI(
openapi_spec=app.openapi(), client=client, name=name, **settings
)
def _convert_to_content(
result: Any,
_process_as_single_item: bool = False,
) -> list[TextContent | ImageContent | EmbeddedResource]:
"""Convert a result to a sequence of content objects."""
if result is None:
return []
if isinstance(result, TextContent | ImageContent | EmbeddedResource):
return [result]
if isinstance(result, Image):
return [result.to_image_content()]
if isinstance(result, list | tuple) and not _process_as_single_item:
# if the result is a list, then it could either be a list of MCP types,
# or a "regular" list that the tool is returning, or a mix of both.
#
# so we extract all the MCP types / images and convert them as individual content elements,
# and aggregate the rest as a single content element
mcp_types = []
other_content = []
for item in result:
if isinstance(item, TextContent | ImageContent | EmbeddedResource | Image):
mcp_types.append(_convert_to_content(item)[0])
else:
other_content.append(item)
if other_content:
other_content = _convert_to_content(
other_content, _process_as_single_item=True
)
return other_content + mcp_types
if not isinstance(result, str):
try:
result = json.dumps(pydantic_core.to_jsonable_python(result))
except Exception:
result = str(result)
return [TextContent(type="text", text=result)]