Nexent Doc

English

简体中文

English

简体中文

Appearance

Model Context Protocol (MCP) ​

🌟 What is MCP? ​

Model Context Protocol (MCP) is an open standard for connecting AI apps to external systems (data, tools, workflows), similar to a "USB-C for AI." It standardizes how hosts (e.g., Claude Desktop, Nexent) discover and call tools/resources exposed by MCP servers.

🧭 What can MCP do? ​

  • Tools: Functions callable by the LLM with user approval
  • Resources: File-like data that clients can read
  • Prompts: Reusable templates shared by servers
  • Works over a simple protocol so hosts can connect to local or remote servers consistently

🌐 Language Support ​

The MCP protocol provides SDKs for multiple programming languages:

  • Python ⭐ (recommended for beginners)
  • TypeScript
  • Java
  • Go
  • Rust
  • Any other language that implements the MCP protocol

We recommend Python because it offers beginner-friendly syntax, rich ecosystem with frameworks like FastMCP, rapid prototyping capabilities, and thousands of mature libraries.

🚀 Quick Start ​

📋 Prerequisites ​

Install FastMCP before you start coding:

bash
pip install fastmcp

📝 Basic Example ​

Create a simple string utility server with FastMCP:

python
from fastmcp import FastMCP

# Create an MCP server instance
mcp = FastMCP(name="String MCP Server")

@mcp.tool(
    name="calculate_string_length",
    description="Calculate the length of a string"
)
def calculate_string_length(text: str) -> int:
    return len(text)

@mcp.tool(
    name="to_uppercase",
    description="Convert text to uppercase"
)
def to_uppercase(text: str) -> str:
    return text.upper()

@mcp.tool(
    name="to_lowercase",
    description="Convert text to lowercase"
)
def to_lowercase(text: str) -> str:
    return text.lower()

if __name__ == "__main__":
    # Start with SSE transport
    mcp.run(transport="sse", port=8000)

🏃 Run the Server ​

Save the code as mcp_server.py and execute:

bash
python mcp_server.py

You should see the server start successfully with the endpoint http://127.0.0.1:8000/sse.

🔌 Integrate with Nexent ​

Once your MCP server is running, connect it to Nexent:

📍 Step 1: Start the MCP Server ​

Keep the server process running and note the endpoint (e.g., http://127.0.0.1:8000/sse).

⚙️ Step 2: Register in Nexent ​

  1. Open the Agent Development page
  2. On the "Select Agent Tools" tab, click MCP Configuration on the right
  3. Enter the server name and server URL
    • ⚠️ Important:
      • Server name must contain only letters and digits (no spaces or symbols)
      • When Nexent runs inside Docker and MCP server runs on the host, replace 127.0.0.1 with host.docker.internal (e.g., http://host.docker.internal:8000)
  4. Click Add to finish registration

🎯 Step 3: Use the MCP Tool ​

During agent creation or editing, the newly registered MCP tool appears in the tool list and can be attached to any agent.

🔧 Advanced Use Cases ​

🌐 Wrap a REST API ​

Expose existing REST APIs as MCP tools:

python
from fastmcp import FastMCP
import requests

mcp = FastMCP("Course Statistics Server")

@mcp.tool(
    name="get_course_statistics",
    description="Get course statistics such as average, max, min, and total students"
)
def get_course_statistics(course_id: str) -> str:
    api_url = "https://your-school-api.com/api/courses/statistics"
    response = requests.get(api_url, params={"course_id": course_id})

    if response.status_code == 200:
        data = response.json()
        stats = data.get("statistics", {})
        return (
            f"Course {course_id} statistics:\n"
            f"Average: {stats.get('average', 'N/A')}\n"
            f"Max: {stats.get('max', 'N/A')}\n"
            f"Min: {stats.get('min', 'N/A')}\n"
            f"Total Students: {stats.get('total_students', 'N/A')}"
        )
    return f"API request failed: {response.status_code}"

if __name__ == "__main__":
    mcp.run(transport="sse", port=8000)

🏢 Wrap an Internal Module ​

Integrate local business logic:

python
from fastmcp import FastMCP
from your_school_module import query_course_statistics

mcp = FastMCP("Course Statistics Server")

@mcp.tool(
    name="get_course_statistics",
    description="Get course statistics such as average, max, min, and total students"
)
def get_course_statistics(course_id: str) -> str:
    try:
        stats = query_course_statistics(course_id)
        return (
            f"Course {course_id} statistics:\n"
            f"Average: {stats.get('average', 'N/A')}\n"
            f"Max: {stats.get('max', 'N/A')}\n"
            f"Min: {stats.get('min', 'N/A')}\n"
            f"Total Students: {stats.get('total_students', 'N/A')}"
        )
    except Exception as exc:
        return f"Failed to query statistics: {exc}"

if __name__ == "__main__":
    mcp.run(transport="sse", port=8000)

✅ Best Practices ​

  • Logging: For stdio transports, avoid stdout logging (no print); log to stderr/files. Logging guidance
  • Documentation: Keep tool docstrings clear; FastMCP derives schema from type hints
  • Error Handling: Handle errors gracefully and return user-friendly text
  • Security: Do not hard-code secrets; load credentials from env/secret managers

📚 Resources ​

🐍 Python ​

🔤 Other Languages ​

📖 Official Documentation ​

🆘 Need Help? ​

If you run into issues while developing MCP servers:

  1. Check the FAQ
  2. Ask questions in GitHub Discussions
  3. Review sample servers on the ModelScope MCP Marketplace