How to quickly implement MCP functionality

[nct_liuhaonan]

How to Quickly Implement MCP Features

Please follow the instructions in this article for configuration.

---

MCP Introduction

MCP (Model Context Protocol) is a mechanism that provides large models with the ability to access external data via APIs.

MCP enables AI to have the ability to “read” and “write” databases, for example:

Read

Based on an order number (or product number), allow the AI to know the information stored in the database and answer the customer with the results.
Example:

Customer asks: “Where is my item?”
→ Search for information in the database based on the “order number” provided by the customer
→ Reply to the customer: “The item you purchased shows [In Transit], please pay attention to receive it.”

---

Write

Based on the order number and the issue mentioned by the customer, automatically write after-sales information into the database.
Example:

Customer asks: “The treadmill I bought isn’t working?”
→ Record the after-sales issue in the database based on the “order number” provided by the customer
→ Reply to the customer: “Our colleague will contact you shortly~”

---

Quickly Experience MCP Features

To help you quickly experience the capabilities of MCP, we provide a test API interface. The data is stored in the following online spreadsheet:

Online Spreadsheet (Click to Enter)

The following example is for “Query (Read)” configuration.
For more advanced features, please refer to:
MCP CRUD Parameter Documentation

---

Configuration Steps

Preparation: Create Access Token (Click to Expand)

1. Open Open Platform → OAuth Clients → Create New
   

   
   
   
   

2. You will be automatically redirected to the page to create a Permanent Token. Record the Token value.
   

   
   
   
   

---

Preparation: Configure Token

1. Go to 3Chat Agent → MCP Authorized Access → Create New
   

   

2. Fill in the information in the pop-up page:

   • Configuration Name
   • Configuration Remarks
   • Configuration Token (Paste the token generated earlier)
     
     

---

Configure AI Task Query Logic

1. In the left menu, go to 3Chat Agent → AI Tasks
2. Create a new AI task, configure it according to the diagram below, and save and run:

Query Order Information
---
When the user asks about order-related information
---
If the customer has not provided an order number in the conversation history: Output "Please provide an order number"
If the customer has provided an order number in the conversation history: Call the "Query Order Information" tool and output the queried order information

Click the plus sign in the top right corner and select the corresponding MCP tool

---

Configure MCP Tool (Configure according to the image)

1. Configuration Name and Description

   The description helps the large model understand when to call this tool

   search_order
   ---
   Query Order Information
   

   

2. Configure Request Parameters

   Parameter descriptions tell the large model how to construct the query request

   https://c2.xinheyun.com/api/open/v2/coze/workflow/run
   ---
   order_id   Customer's order number
   ---
   {
     "workflowId": "7555717286728908834",
     "parameters": {
       "base_url": "https://3chat-ai.feishu.cn/base/Ct31bmyqgaquzxsjGG0cETeRnzc?table=tbl4tFYj2pkMk9ub&view=vewfieQ4B1",
       "fields": {},
       "order_id": "{{order_id}}",
       "purpose": "read"
     }
   }
   

   

3. Configure Parameter Values and Test the Interface

   You can add test order numbers in the online spreadsheet yourself

   order_id   P1234567896112222222
   

   

4. Clean Return Parameters

   • Click the button to disable “Accessible” to reduce redundant data
   • Keep only the parameter section marked in the image
     
     

5. Click Save

---

Final Test Results