It provides:

• FastMCPProvider with FastMCP request context objects for dishka.
• setup_dishka() middleware that opens a Scope.REQUEST container per tool/resource/prompt call.
• @inject support for parameters marked as FromDishka[T].

Usage

from dishka import Provider, Scope, make_async_container, provide
from fastmcp import FastMCP
from fastmcp_dishka import FastMCPProvider, FromDishka, inject, setup_dishka


class GreetingService:
    def greet(self, name: str) -> str:
        return f"Hello, {name}!"


class AppProvider(Provider):
    greeting = provide(GreetingService, scope=Scope.REQUEST)


mcp = FastMCP("GreetMCP")
container = make_async_container(AppProvider(), FastMCPProvider())
setup_dishka(container, mcp)


@mcp.tool
@inject
async def greet(name: str, service: FromDishka[GreetingService]) -> str:
    return service.greet(name)

Place @inject below FastMCP decorators so FastMCP registers the wrapped function signature.

@mcp.prompt defines a reusable prompt template exposed to MCP clients. A prompt does not perform an action like a tool; it returns text or chat messages that a client can insert into an LLM conversation.

@mcp.prompt
@inject
async def welcome_prompt(
    name: str,
    service: FromDishka[GreetingService],
) -> str:
    return f"Write a short welcome message for: {service.greet(name)}"

See examples/fastmcp_app.py for a complete FastMCP server with a tool, resource, prompt, and Dishka providers.

Development

Install development dependencies:

uv sync --dev

Run tests with coverage:

make test

Run linting and formatting:

make lint
make format

Install pre-commit hooks:

make pre-commit-install