convez.ai
MCP Server
apisani1

mcp-multi-server

apisani1/mcp-multi-server

Ver no GitHub
InicialPluginScore · de 0 a 100

Instalação

$npx convez-ai apisani1/mcp-multi-server

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 1 semana
Versão
v1.2.5
Issues abertas
1
Contribuidores
2

README

A tradução para o português deste README ainda está na fila.

MCP Multi-Server

Python 3.10+ License: MIT

A Python library for managing connections to multiple Model Context Protocol (MCP) servers. This library provides a unified interface for discovering, aggregating, and routing capabilities (tools, resources, prompts) across multiple MCP servers.

Features

  • Multi-Server Management: Connect to and manage multiple MCP servers simultaneously
  • Automatic Capability Discovery: Discover tools, resources, prompts, and templates from all connected servers
  • Intelligent Routing: Automatically route tool calls, resource reads, and prompt retrievals to the correct server
  • Namespace Support: Use namespaced URIs for unambiguous resource routing
  • Collision Detection: Detect and warn about duplicate tool or prompt names across servers
  • Async Context Manager: Clean resource management with Python's async context managers
  • Sync & Async Support: Both async (MultiServerClient) and synchronous (SyncMultiServerClient) interfaces

Installation

pip install mcp-multi-server

Or with Poetry:

poetry add mcp-multi-server

Optional Dependencies

For OpenAI integration:

pip install mcp-multi-server[openai]

For running examples:

pip install mcp-multi-server[examples]

Quick Start

1. Create a Server Configuration File

Create a mcp_servers.json file defining your MCP servers:

{
  "mcpServers": {
    "filesystem": {
      "command": "python",
      "args": ["-m", "my_servers.filesystem_server"]
    },
    "database": {
      "command": "python",
      "args": ["-m", "my_servers.database_server"]
    }
  }
}

2. Use the Multi-Server Client

import asyncio
from mcp_multi_server import MultiServerClient

async def main():
    # Using context manager (recommended)
    async with MultiServerClient.from_config("mcp_servers.json") as client:
        # List all available tools from all servers
        tools = client.list_tools()
        print(f"Found {len(tools.tools)} tools")

        # Call a tool (automatically routed to the correct server)
        result = await client.call_tool(
            "read_file",
            {"path": "/path/to/file.txt"}
        )

        # List all resources with namespaced URIs
        resources = client.list_resources()

        # Read a resource (auto-routing via namespace)
        content = await client.read_resource(resources.resources[0].uri)

        # Get a prompt
        prompt = await client.get_prompt("code_review", {"language": "python"})

asyncio.run(main())

3. Programmatic Configuration

You can also configure servers programmatically without a JSON file:

from mcp_multi_server import MultiServerClient, MCPServersConfig, ServerConfig

config = MCPServersConfig(mcpServers={
    "my_server": ServerConfig(
        command="python",
        args=["-m", "my_package.my_server"]
    )
})

async with MultiServerClient.from_dict(config.model_dump()) as client:
    tools = client.list_tools()
    # ...

4. Synchronous Client

For non-async code, use the synchronous wrapper:

from mcp_multi_server import SyncMultiServerClient

# Using context manager (recommended)
with SyncMultiServerClient.from_config("mcp_servers.json") as client:
    tools = client.list_tools()
    result = client.call_tool("read_file", {"path": "/path/to/file.txt"})
    resources = client.list_resources()

# Or with programmatic configuration
config = {"mcpServers": {"my_server": {"command": "python", "args": ["-m", "my_server"]}}}
with SyncMultiServerClient.from_dict(config) as client:
    tools = client.list_tools()

The sync client runs a background event loop thread and provides the same API as the async client, with optional timeout parameters on blocking methods.

Examples

The repository includes comprehensive examples demonstrating various use cases. See the examples directory for:

  • Example MCP server implementations (tools, resources, prompts)
  • Example chat client showing usage patterns
  • Full client with OpenAI integration

API Reference

MultiServerClient

Main class for managing multiple MCP servers.

Class Methods:

  • from_config(config_path: str) - Create client from JSON config file
  • from_dict(config_dict: Dict) - Create client from configuration dictionary

Instance Methods:

  • connect_all(stack: AsyncExitStack) - Connect to all configured servers
  • list_tools() - Get all tools from all servers
  • list_prompts() - Get all prompts from all servers
  • list_resources(use_namespace: bool = True) - Get all resources
  • list_resource_templates(use_namespace: bool = True) - Get all resource templates
  • call_tool(name, arguments, server_name=None) - Call a tool
  • read_resource(uri, server_name=None) - Read a resource
  • get_prompt(name, arguments=None, server_name=None) - Get a prompt

SyncMultiServerClient

Synchronous wrapper for non-async code. Same API as MultiServerClient.

Class Methods:

  • from_config(config_path: str) - Create client from JSON config file
  • from_dict(config_dict: Dict) - Create client from configuration dictionary

Instance Methods:

  • list_tools() - Get all tools from all servers
  • list_prompts() - Get all prompts from all servers
  • list_resources(use_namespace: bool = True) - Get all resources
  • list_resource_templates(use_namespace: bool = True) - Get all resource templates
  • call_tool(name, arguments, timeout=None, server_name=None) - Call a tool
  • read_resource(uri, timeout=None, server_name=None) - Read a resource
  • get_prompt(name, arguments=None, timeout=None, server_name=None) - Get a prompt
  • shutdown() - Explicitly shutdown the client

Utility Functions

  • print_capabilities_summary(client) - Print client discovered capabilities
  • mcp_tools_to_openai_format(tools) - Convert MCP tools to OpenAI function format
  • format_namespace_uri(server_name, uri) - Create namespaced URI
  • parse_namespace_uri(uri) - Parse namespaced URI
  • extract_template_variables(template) - Extract variables from URI template
  • substitute_template_variables(template, variables) - Substitute template variables

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Links

Ver o original no GitHub

Relacionados