MCP Serializer

MCP Serializer

Registers MCP features (tools, prompts, resources) and serializes JSON-RPC requests into response data.

Category
访问服务器

README

MCP Serializer

Register MCP features and serialize JSON-RPC requests into response data.

MCP Serializer is a Python library that handles the Model Context Protocol (MCP) serialization layer for your server. It provides:

  • Feature Registration: Define tools, prompts, and resources using simple decorators and functions
  • Request Processing: Process incoming JSON-RPC requests and get properly formatted MCP response data

This library focuses on the protocol implementation - you register what your server can do, and the serializer handles converting requests into responses. You're responsible for the transport layer (HTTP, stdio, or any other communication method) and wiring up the request/response cycle with your chosen framework.

Installation

pip install mcp-serializer

Feature Registration

A registry instance is needed to register tools, prompts, and resources.

from mcp_serializer.registry import MCPRegistry

registry = MCPRegistry()

<br>

1. Registering Resources

Register a file as a resource. The file parameter is required and can be a file path or a file object.

registry.add_file_resource(
    file="/path/to/file.json",
    title="File Resource Title",
    description="This is a file resource",
)

The file will be automatically converted to text or binary content based on the mime type parsed from the file extension. You can supply a custom URI using the uri parameter, otherwise it will be generated from the file path.

Register an HTTP resource by providing a URI. The URI is the only required parameter.

registry.add_http_resource(
    uri="https://example.com/image.png",
    title="Profile Image",
    description="This image can be used as a profile image for a user.",
    mime_type="image/png",
    size=1024,
)

The mime type will be taken from the mime_type parameter or will be automatically determined from the URI extension.

For complex cases, register a resource using a function that returns a ResourceResult object. Functions with parameters create resource templates with URI placeholders.

from mcp_serializer.results import ResourceResult

@registry.resource(uri="resource/weather/")
def get_weather(city: str):
    """Get weather information for a given city

    Returns weather data for the specified city including temperature and conditions.
    """
    result = ResourceResult()

    # Add text content directly
    result.add_text_content(
        f"today is cold in {city}",
        mime_type="text/plain",
        title="indicated cold/hot in the city",
    )

    # Add file as resource (converted to text/binary based on file extension)
    result.add_file("/path/to/weather_file.json")

    return result

This creates a resource template with URI resource/weather/{city}. The function name becomes the resource name. The first line of the docstring becomes the title, and the rest becomes the description.

<br>

2. Registering Tools

A function can be registered as a tool by using the @registry.tool() decorator.

@registry.tool()
def add_weather_data(
    city: str, temperature: float, wind_speed: float = None, humidity: float = None
):
    """Add weather data for a given city

    This tool saves weather data to the database for a specific city.

    Args:
        city: The name of the city
        temperature: The temperature in Fahrenheit
        wind_speed: The wind speed in mph
        humidity: The humidity percentage
    """
    # Your logic to save weather data goes here
    return f"Weather data for {city} added successfully"

The function name becomes the tool name. The docstring is parsed as Google-style docstring:

  • First line(s) followed by a new line becomes the title
  • Description is the text until the Args section
  • Parameter description from Args section is used as description for the tool's input schema

For structured data, return a Pydantic BaseModel. The model must be specified as the return type.

from pydantic import BaseModel

class WeatherReport(BaseModel):
    temperature: float
    condition: str
    humidity: float

@registry.tool()
def get_current_weather(city: str) -> WeatherReport:
    """Get current weather for a city

    Retrieves real-time weather information including temperature, condition, and humidity.

    Args:
        city: The name of the city
    """
    return WeatherReport(temperature=72.5, condition="sunny", humidity=65.0)

The Pydantic model will be automatically converted to an output schema in the tool definition.

For complex responses with multiple content types, return a ToolsResult object.

from mcp_serializer.results import ToolsResult

@registry.tool()
def get_weather_forecast(city: str, days: int = 3) -> ToolsResult:
    """Get weather forecast for a city

    Provides a detailed weather forecast with text, data files, and resource links.

    Args:
        city: The name of the city
        days: Number of days to forecast (default: 3)
    """
    result = ToolsResult()

    # Add text content
    result.add_text_content(
        f"{days}-day forecast for {city}: Mostly sunny with temperatures around 70�F"
    )

    # Add embedded file resource
    result.add_file("/path/to/file.json")

    # Add resource link to an existing resource
    result.add_resource_link(uri="resource/weather/{city}", registry=registry)

    return result

ToolsResult also has add_image_content and add_audio_content functions to add contents manually.

<br>

3. Registering Prompts

Register a text prompt directly.

registry.add_text_prompt(
    name="greeting",
    text="Hello, how are you?",
    title="Greeting a user and asking common questions",
    role="user",
)

The name and text are mandatory. Role can be "user" or "assistant" and defaults to "user".

Register a text prompt from a file.

registry.add_file_prompt(
    name="create_user",
    description="Create a new user for any purpose.",
    file="/path/to/prompt.txt",
    role="user",
)

The file can be a file path or a file object. You can supply title and description as parameters for these registration methods. These are useful for the client to understand the prompt.

A function can be registered to create a prompt. It can return a string, a tuple (text, role), or a PromptsResult object.

from mcp_serializer.results import PromptsResult

@registry.prompt()
def greeting_prompt(name: str):
    """Greeting prompt

    This prompt helps to greet a user.
    """
    result = PromptsResult()

    # Create a text prompt content (same as returning a string)
    result.add_text(f"Hello, {name}! How are you?", role=PromptsResult.Roles.USER)

    # Create a text prompt from a file
    result.add_file_message("/path/to/message.md")

    # Embed a resource with your prompt
    result.add_file_resource("/path/to/resource.json")

    return result

This creates a prompt with the name "greeting_prompt". The function docstring is parsed as Google-style:

  • Title will be the first lines of docstring followed by a new line
  • The description will be the rest of the docstring

<br>

Initializer

An initializer is needed to return the MCP initialization data. You need to add your features to present at initialization time.

from mcp_serializer.initializer import MCPInitializer

initializer = MCPInitializer()
initializer.add_server_info("My MCP Server", "1.0.0", "A title of the server.")
initializer.add_prompt()
initializer.add_resources()
initializer.add_tools()

Custom Initializer

You can create your own initializer by inheriting from the MCPInitializer class. You can override the build_result method to initialize MCP client parameters.

class MyInitializer(MCPInitializer):
    def build_result(self, client_params: dict):
        self.protocol_version = client_params.get(
            "protocolVersion", self.protocol_version
        )
        return super().build_result(client_params)

<br>

Creating and Using the Serializer

from mcp_serializer.serializers import MCPSerializer

serializer = MCPSerializer(initializer=initializer, registry=registry, page_size=10)

The page_size is the number of items to return in a single page for listing features.

Processing Requests

The process_request method takes a JSON-RPC request and returns a ResponseContext object. The ResponseContext provides access to the response data and maintains a history of all request-response interactions.

response_context = serializer.process_request(
    request_data={
        "jsonrpc": "2.0",
        "id": 1,
        "method": "initialize",
        "params": {
            "protocolVersion": "2024-11-05",
            "clientInfo": {"name": "test-client", "version": "1.0.0"},
        },
    }
)

# Access the response data as a dictionary (ready to send to client)
response_dict = response_context.response_data

# Access detailed information through history
for entry in response_context.history:
    # Each entry contains:
    # - entry.response: Pydantic response object (JsonRpcSuccessResponse/JsonRpcErrorResponse)
    # - entry.request: Pydantic request object (JsonRpcRequest)
    # - entry.data: Response as dictionary
    # - entry.is_error: Boolean indicating if this is an error
    # - entry.is_notification: Boolean indicating if this is a notification
    
    if entry.is_error:
        print("Error:", entry.data["error"])
    elif entry.is_notification:
        print("Notification - no response to send")
    else:
        print("Success:", entry.data["result"])

The request_data parameter can be a dict or a JSON string.

ResponseContext Properties:

  • response_data: The response as a dictionary or list of dictionaries (for batch requests), ready to be serialized to JSON and sent to the client
  • history: List of ResponseEntry objects, one per request processed

ResponseEntry Properties (accessible via response_context.history[i]):

  • response: The Pydantic BaseModel response object (JsonRpcSuccessResponse, JsonRpcErrorResponse, or None for notifications)
    • JsonRpcSuccessResponse properties: jsonrpc, id, result
    • JsonRpcErrorResponse properties: jsonrpc, id, error
  • request: The Pydantic BaseModel request object (JsonRpcRequest)
    • JsonRpcRequest properties: jsonrpc, id, method, params
  • data: The response as a dictionary (same as the corresponding item in response_data)
  • is_error: Boolean property indicating if this is an error response
  • is_notification: Boolean property indicating if this was a notification (no response needed)

<br>

Examples

The following examples demonstrate the complete request-response cycle using the features registered in this documentation. If you follow the registration steps outlined above and create a serializer with the same configuration, you can expect these exact request-response patterns.

Initialization

Request:

{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "initialize",
    "params": {
        "protocolVersion": "2024-11-05",
        "capabilities": {},
        "clientInfo": {"name": "test-client", "version": "1.0.0"}
    }
}

Response:

{
    "jsonrpc": "2.0",
    "id": 1,
    "result": {
        "protocolVersion": "2024-11-05",
        "serverInfo": {
            "name": "My MCP Server",
            "title": "A title of the server.",
            "version": "1.0.0"
        },
        "capabilities": {
            "prompts": {"listChanged": false},
            "resources": {"subscribe": false, "listChanged": false},
            "tools": {"listChanged": false}
        }
    }
}

Tools

List Tools

Request:

{
    "jsonrpc": "2.0",
    "id": 2,
    "method": "tools/list",
    "params": {}
}

Response:

{
    "jsonrpc": "2.0",
    "id": 2,
    "result": {
        "tools": [
            {
                "name": "add_weather_data",
                "title": "Add weather data for a given city",
                "description": "This tool saves weather data to the database for a specific city.",
                "inputSchema": {
                    "type": "object",
                    "properties": {
                        "city": {
                            "type": "string",
                            "description": "The name of the city"
                        },
                        "temperature": {
                            "type": "number",
                            "description": "The temperature in Fahrenheit"
                        },
                        "wind_speed": {
                            "type": "number",
                            "description": "The wind speed in mph"
                        },
                        "humidity": {
                            "type": "number",
                            "description": "The humidity percentage"
                        }
                    },
                    "required": ["city", "temperature"]
                }
            },
            {
                "name": "get_current_weather",
                "title": "Get current weather for a city",
                "description": "Retrieves real-time weather information including temperature, condition, and humidity.",
                "inputSchema": {
                    "type": "object",
                    "properties": {
                        "city": {
                            "type": "string",
                            "description": "The name of the city"
                        }
                    },
                    "required": ["city"]
                },
                "outputSchema": {
                    "type": "object",
                    "properties": {
                        "temperature": {"type": "number"},
                        "condition": {"type": "string"},
                        "humidity": {"type": "number"}
                    },
                    "required": ["temperature", "condition", "humidity"]
                }
            },
            {
                "name": "get_weather_forecast",
                "title": "Get weather forecast for a city",
                "description": "Provides a detailed weather forecast with text, data files, and resource links.",
                "inputSchema": {
                    "type": "object",
                    "properties": {
                        "city": {
                            "type": "string",
                            "description": "The name of the city"
                        },
                        "days": {
                            "type": "integer",
                            "description": "Number of days to forecast (default: 3)"
                        }
                    },
                    "required": ["city"]
                }
            }
        ]
    }
}

Call Tool - Simple String Return

Request:

{
    "jsonrpc": "2.0",
    "id": 3,
    "method": "tools/call",
    "params": {
        "name": "add_weather_data",
        "arguments": {
            "city": "London",
            "temperature": 15.5,
            "wind_speed": 10.0,
            "humidity": 80.0
        }
    }
}

Response:

{
    "jsonrpc": "2.0",
    "id": 3,
    "result": {
        "content": [
            {"type": "text", "text": "Weather data for London added successfully"}
        ]
    }
}

Call Tool - Pydantic Model Return

Request:

{
    "jsonrpc": "2.0",
    "id": 3,
    "method": "tools/call",
    "params": {
        "name": "get_current_weather",
        "arguments": {
            "city": "Paris"
        }
    }
}

Response:

{
    "jsonrpc": "2.0",
    "id": 3,
    "result": {
        "content": [],
        "structuredContent": {
            "condition": "sunny",
            "humidity": 65.0,
            "temperature": 72.5
        }
    }
}

Call Tool - ToolsResult with Multiple Content Types

Request:

{
    "jsonrpc": "2.0",
    "id": 3,
    "method": "tools/call",
    "params": {
        "name": "get_weather_forecast",
        "arguments": {"city": "Tokyo", "days": 5}
    }
}

Response:

{
    "jsonrpc": "2.0",
    "id": 3,
    "result": {
        "content": [
            {
                "type": "text",
                "text": "5-day forecast for Tokyo: Mostly sunny with temperatures around 70�F"
            },
            {
                "type": "resource",
                "resource": {
                    "uri": "file:///path/to/file.json",
                    "mimeType": "application/json",
                    "name": "file.json",
                    "text": "{\"test\": \"data\", \"version\": \"1.0\"}"
                }
            },
            {
                "type": "resource_link",
                "uri": "resource/weather/{city}",
                "name": "get_weather",
                "title": "Get weather information for a given city",
                "description": "Returns weather data for the specified city including temperature and conditions."
            }
        ]
    }
}

Prompts

List Prompts

Request:

{
    "jsonrpc": "2.0",
    "id": 4,
    "method": "prompts/list",
    "params": {}
}

Response:

{
    "jsonrpc": "2.0",
    "id": 4,
    "result": {
        "prompts": [
            {
                "name": "create_user",
                "description": "Create a new user for any purpose."
            },
            {
                "name": "greeting",
                "title": "Greeting a user and asking common questions"
            },
            {
                "name": "greeting_prompt",
                "title": "Greeting prompt",
                "description": "This prompt helps to greet a user.",
                "arguments": [{"name": "name", "type": "string", "required": true}]
            }
        ]
    }
}

Get Simple Prompt

Request:

{
    "jsonrpc": "2.0",
    "id": 5,
    "method": "prompts/get",
    "params": {"name": "greeting"}
}

Response:

{
    "jsonrpc": "2.0",
    "id": 5,
    "result": {
        "messages": [
            {
                "role": "user",
                "content": {"type": "text", "text": "Hello, how are you?"}
            }
        ]
    }
}

Get Prompt with Arguments

Request:

{
    "jsonrpc": "2.0",
    "id": 6,
    "method": "prompts/get",
    "params": {"name": "greeting_prompt", "arguments": {"name": "Alice"}}
}

Response:

{
    "jsonrpc": "2.0",
    "id": 6,
    "result": {
        "description": "This prompt helps to greet a user.",
        "messages": [
            {
                "role": "user",
                "content": {"type": "text", "text": "Hello, Alice! How are you?"}
            },
            {
                "role": "user",
                "content": {
                    "type": "text",
                    "text": "You are a helpful assistant that can help with any questions.",
                    "mimeType": "text/markdown"
                }
            },
            {
                "role": "user",
                "content": {
                    "type": "resource",
                    "resource": {
                        "uri": "file:///path/to/file.json",
                        "text": "File Resource Title",
                        "mimeType": "application/json",
                        "name": "file.json",
                        "text": "{\"test\": \"data\", \"version\": \"1.0\"}"
                    }
                }
            }
        ]
    }
}

Resources

List Resources

Request:

{
    "jsonrpc": "2.0",
    "id": 7,
    "method": "resources/list",
    "params": {}
}

Response:

{
    "jsonrpc": "2.0",
    "id": 7,
    "result": {
        "resources": [
            {
                "uri": "file:///path/to/file.json",
                "name": "file.json",
                "title": "File Resource Title",
                "description": "This is a file resource",
                "size": 34,
                "mimeType": "application/json"
            },
            {
                "uri": "https://example.com/image.png",
                "title": "Profile Image",
                "description": "This image can be used as a profile image for a user.",
                "size": 1024,
                "mimeType": "image/png"
            }
        ]
    }
}

List Resource Templates

Request:

{
    "jsonrpc": "2.0",
    "id": 8,
    "method": "resources/templates/list",
    "params": {}
}

Response:

{
    "jsonrpc": "2.0",
    "id": 8,
    "result": {
        "resourceTemplates": [
            {
                "uri": "resource/weather/{city}",
                "name": "get_weather",
                "title": "Get weather information for a given city",
                "description": "Returns weather data for the specified city including temperature and conditions."
            }
        ]
    }
}

Read File Resource

Request:

{
    "jsonrpc": "2.0",
    "id": 10,
    "method": "resources/read",
    "params": {"uri": "file:///path/to/file.json"}
}

Response:

{
    "jsonrpc": "2.0",
    "id": 10,
    "result": {
        "contents": [
            {
                "uri": "file:///path/to/file.json",
                "name": "file.json",
                "title": "File Resource Title",
                "mimeType": "application/json",
                "text": "{\"test\": \"data\", \"version\": \"1.0\"}"
            }
        ]
    }
}

Read Resource Template with Parameters

Request:

{
    "jsonrpc": "2.0",
    "id": 12,
    "method": "resources/read",
    "params": {"uri": "resource/weather/Paris"}
}

Response:

{
    "jsonrpc": "2.0",
    "id": 12,
    "result": {
        "contents": [
            {
                "mimeType": "text/plain",
                "name": "get_weather",
                "title": "indicated cold/hot in the city",
                "text": "today is cold in Paris",
                "uri": "resource/weather"
            },
            {
                "uri": "file:///path/to/weather_file.json",
                "mimeType": "application/json",
                "name": "weather_file.json",
                "text": "{\"temperature\": 72, \"condition\": \"sunny\"}"
            }
        ]
    }
}

Error Handling

Request:

{
    "jsonrpc": "2.0",
    "id": 12,
    "method": "invalid/method",
    "params": {}
}

Response:

{
    "jsonrpc": "2.0",
    "id": 12,
    "error": {
        "code": -32601,
        "message": "Method not found",
        "data": {"method": "invalid/method"}
    }
}

Batch Requests

Request:

[
    {"jsonrpc": "2.0", "id": 13, "method": "tools/list", "params": {}},
    {"jsonrpc": "2.0", "id": 14, "method": "prompts/list", "params": {}},
    {"jsonrpc": "2.0", "id": 15, "method": "resources/list", "params": {}}
]

Response:

[
    {"jsonrpc": "2.0", "id": 13, "result": {...}},
    {"jsonrpc": "2.0", "id": 14, "result": {...}},
    {"jsonrpc": "2.0", "id": 15, "result": {...}}
]

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选