3.8 Plugins & MCP Tools

Pop’s plugin system (Plugins) and MCP tools (Model Context Protocol Tools) give AI true execution ability, upgrading it from a Q&A assistant to a fully capable AI agent.

This chapter introduces Pop’s plugin system, supported MCP tools, extension methods, and how tools are triggered during AI conversations.

🚀 What Are MCP Tools?

MCP (Model Context Protocol) is a new standard proposed by OpenAI to provide unified tool‑calling capabilities for AI.

In simple terms:

A multimodal model → issues tool‑invocation instructions → the client executes real tasks and returns results.

Pop fully supports the MCP specification, enabling:

Reading files
Modifying projects
Database queries
Running local scripts
Executing HTTP/network requests
Calling external APIs
Running complex workflows

This gives AI operational capability—it no longer only talks; it can actually do things.

🔌 Components of Pop’s Tool System

Pop provides two categories of tool capabilities:

1) Built‑in Tools

Available out‑of‑the‑box, including:

File read/write
JSON parsing/formatting
Image compression & conversion
Text analysis tools
Code formatting
Workflow execution
Document generation (PDF, Excel, PPT)

2) External MCP Tools

You can install any MCP‑compatible tool, such as:

Official OpenAI tools
FS file‑management tools
Web/HTTP request tools
Database tools (SQLite, MySQL)
Shell executor tools
Git tools
Python runner
Node.js runner

Any tool that follows the MCP protocol is automatically supported.

🛠 Typical Tool Invocation Flow

When you ask certain questions, AI may automatically:

Decide whether tool invocation is needed
Construct a tool‑call request
Pop executes the real task
Returns results to the model
The model continues and provides the final answer

Example:

“Count how many PNG files are in the current folder.”

AI will automatically:

Call fs.readdir
Filter .png
Return the count
Continue explaining

You don’t need to know any backend logic—Pop handles everything.

🧩 Common Tool Examples

● File System (FS)

List files
Read files
Write files
Create/delete folders

● Shell Tools

Execute system commands such as:

ls -lah
npm run build
python3 script.py

● HTTP Tools

Make network requests:

GET https://api.example.com/data
POST with JSON payload

● Code Tools

Formatting
Linting
Auto‑fixing issues

● Document‑generation Tools

PDF / Docx / Excel
Auto‑create PPT slides

These are heavily used inside workflows.

🔗 Managing MCP Tools in Pop

In the MCP Tools Center, you can:

Add MCP servers
Enable/disable each tool
View available capabilities
Refresh tool lists
Test tool functionality

Tool sources include:

Local services (Node.js / Python)
Official tool repositories
Third‑party libraries
Custom‑developed tools

🧪 Developing Your Own MCP Tool

Pop provides templates for MCP development.

Requirements:

Node.js or Python
MCP SDK
Tool configuration
Implementation logic

Example (Python):

from mcp import tool

@tool()
def hello(name: str):
    return f"Hello, {name}!"

After registration, Pop automatically detects and loads the tool.

🤖 How AI Automatically Selects Tools

Pop includes an intelligent decision system:

When the user issues an actionable command
AI evaluates if a tool is required
If needed → constructs MCP tool‑call JSON
Otherwise → responds normally

Example:

“Create a project structure with README.md.”

AI will automatically:

mkdir
writeFile
Generate initial project structure

Everything runs transparently.

🔒 Tool Permissions & Safety

Pop enforces strict security control:

Tools are disabled by default
High‑risk tools (e.g., Shell) must be manually enabled
Permission prompts appear before tool invocation
All actions are logged
Sandbox isolation is applied

Suitable for enterprise environments.

📌 Tool + Workflow Integration

Tools also serve as execution engines for workflows:

Code generation → write files
Excel generation → auto‑save
PDF processing → pass to next node
HTTP requests → feed downstream steps

Tool capabilities are the core of Pop’s workflow execution.

📌 Summary

Plugins and MCP tools give Pop real execution power, enabling:

Local file operations
Document generation
Network requests
Codebase management
Automation tasks
External system integration
Workflow execution

This transforms Pop from a simple Q&A assistant into a true AI Operating System (AI‑OS).