FastMCP Extensions Library

Warnings

• breaking Version 0.3.0 of `fastmcp-extensions` upgrades its core dependency `fastmcp` from v2 to v3. Projects upgrading to `fastmcp-extensions` v0.3.0 must also ensure their `fastmcp` installation is v3 or higher and that their server code is compatible with `FastMCP` v3's API changes.
  Fix: Upgrade `fastmcp` to a compatible v3.x version (e.g., `pip install 'fastmcp>=3.0.0'`) and review FastMCP v3 migration guides for any breaking changes in your server implementation.
• gotcha `fastmcp-extensions` is described as an 'unofficial extension library'. This may imply potential for API instability, differing design philosophies, or slower updates compared to official `FastMCP` core developments. Relying heavily on specific extension implementations may introduce tighter coupling.
  Fix: Regularly review the `fastmcp-extensions` GitHub repository for updates and breaking changes. Be prepared to adapt your code if upstream `FastMCP` changes impact the extension library. Consider the long-term maintenance implications for critical applications.
• gotcha The `mcp_server()` helper includes 'built-in server info and credential resolution'. Incorrect configuration or missing environment variables for credentials can lead to authentication failures or security vulnerabilities if defaults are inadvertently used.
  Fix: Thoroughly understand how `mcp_server()` resolves credentials (e.g., via specific environment variables or configuration files). Ensure all required credentials are securely provided and documented for your deployment environment.
• gotcha The 'dynamic tool filtering with intelligent default filtering policy' feature can unexpectedly hide or expose tools if the filtering logic is not fully understood. This could lead to functional gaps for LLM agents or unintended access to sensitive tools.
  Fix: Carefully review the default filtering policies and customize them as needed for your specific use case. Implement robust testing to verify that only the intended tools are available under various contexts and user roles.

Install

• pip install fastmcp-extensions Install latest version

Imports

• mcp_server

  from fastmcp_extensions.server import mcp_server

  Based on release notes; specific modules for other extensions like dynamic tool filtering may vary.
• FastMCP

  from fastmcp import FastMCP

  The core FastMCP object, which fastmcp-extensions extends.

Quickstart

This quickstart demonstrates how to initialize a FastMCP server and conceptually integrate features provided by `fastmcp-extensions`, such as the `mcp_server` helper for server setup and credential resolution, and a placeholder for dynamic tool filtering. Due to the lack of specific examples in public documentation, some imports and usages are illustrative based on the library's stated features. Users should consult the library's source for precise API details.

import os
from fastmcp import FastMCP
# Assuming mcp_server helper is directly exposed in a 'server' module
# Exact import path for specific extensions may vary based on internal structure.
try:
    from fastmcp_extensions.server import mcp_server
    from fastmcp_extensions.filters import dynamic_filter_policy # Hypothetical import
except ImportError:
    print("Warning: fastmcp-extensions specific imports not found. Using FastMCP core only.")
    mcp_server = None # Placeholder
    dynamic_filter_policy = None # Placeholder

def run_extended_mcp_server():
    mcp = FastMCP(name="MyExtendedMCPServer")

    @mcp.tool
    def hello(name: str = "World") -> str:
        """Greets the given name."""
        return f"Hello, {name}! This is an extended FastMCP server."

    # Example of integrating an extension, if 'mcp_server' exists
    if mcp_server:
        # The mcp_server helper likely takes a FastMCP instance and applies configurations.
        # This is a hypothetical usage based on 'built-in server info and credential resolution'.
        print("Applying fastmcp-extensions mcp_server helper...")
        # mcp_server might take parameters for credential resolution, e.g., via env vars.
        # For example, os.environ.get('MCP_SERVER_SECRET_KEY', 'default_key')
        configured_mcp = mcp_server(mcp, credentials_env_var='FASTMCP_API_KEY')
    else:
        configured_mcp = mcp

    # If dynamic_filter_policy exists, it would likely be applied to the server
    if dynamic_filter_policy:
        print("Applying dynamic tool filtering policy...")
        # configured_mcp.add_filter(dynamic_filter_policy())

    print(f"Running {configured_mcp.name} with stdio transport...")
    # In a real scenario, you'd run the server (e.g., in a __main__ block)
    # configured_mcp.run(transport="stdio")
    print("Server setup complete (not actually running in this snippet).")
    print("Available tools on the server:")
    for tool_name, tool_obj in configured_mcp._tools.items():
        print(f"- {tool_name}: {tool_obj.description}")

if __name__ == "__main__":
    # Set a dummy environment variable if needed for hypothetical credential resolution
    os.environ['FASTMCP_API_KEY'] = 'fake_api_key_123'
    run_extended_mcp_server()