system-settings/MCP Services

8.6 MCP Services (Model Context Protocol)

MCP (Model Context Protocol) is Pop's "tool extension system,"
allowing Pop to directly call external services, file systems, databases, browser environments, or third-party APIs,
making AI no longer a "pure chat model," but an intelligent agent with real tool usage capabilities.

For example:

  • Let AI automatically read local files
  • Let AI use browser tools to scrape web pages
  • Let AI operate databases
  • Let AI use the command line
  • Let AI call third-party APIs (weather, search, translation, etc.)
  • Expose any of your own APIs to Pop through custom MCP

MCP is one of the most important capabilities for advanced Pop users. This section will help you master its configuration.

1. MCP Services Overview

The core functions of MCP services:

Function Description
Tool Management (Tools) Control which tools are available to AI, such as Filesystem, Browser, etc.
Service Management (Servers) Control service start, stop, restart
Permission Management Decide whether each tool can access local resources or sensitive interfaces
Configuration Management Set service paths, environment variables, ports, etc.
Debugging Features View service logs, communication status, capability registration

Pop has a built-in MCP management panel to visually manage the entire MCP ecosystem.

2. Installed MCP Servers Management

Pop automatically scans MCP services on your machine or included with Pop and displays them uniformly in "System Settings → MCP Services."

You can see:

  • Service name
  • Service type (filesystem / browser / custom)
  • Running status (running / stopped)
  • Auto-start settings
  • Service log button
  • Tool list

Supported operations

Operation Description
Start Service Manually start the MCP service process
Stop Service Stop the current MCP service
Auto-start Automatically start MCP service when Pop starts
Refresh Tool List Reload the tool capabilities provided by the service
View Logs View service stdout/stderr for debugging

3. Common Built-in MCP Services

Pop by default supports the following official MCP services (depending on installation):

1. Filesystem (Local File System)

Allows AI to:

  • Read file contents
  • Search directories
  • List file trees
  • Create/edit files (risky features can be disabled)

🚨 For security reasons, only the Pop working directory or user-specified directories are open by default.

2. Browser (Web Scraping Tool)

Allows AI to:

  • Access any URL
  • Scrape webpage content (HTML / text)
  • Perform DOM queries
  • Obtain metadata (title, description, etc.)

Suitable for:

  • Knowledge retrieval
  • Page summarization
  • Webpage structure analysis

3. Python / Node Sandbox (Optional)

Allows AI to:

  • Execute Python / Node code in a secure sandbox
  • Data processing
  • Call external APIs (restricted)
  • Plotting/calculation, etc.

4. Service Directory and Permissions

Each MCP service may include some sensitive operations, so Pop provides clear permission configurations.

Common permissions include:

Permission Type Purpose
File Read Permission Whether AI is allowed to read local files
File Write Permission Whether AI is allowed to write files (strongly advised to enable cautiously)
Network Access Permission Whether external HTTP requests are allowed
Command Execution Permission Whether calling system commands is allowed
Custom Directory Whitelist Restrict AI access to specific directories
Sandbox Isolation Mode Restrict the service to run only within Pop's sandbox

You can configure permissions for each tool individually, and Pop will prompt and intercept when AI requests tool capabilities.

5. Custom MCP Services

You can develop your own MCP service and let Pop automatically recognize and use it.

Developing an MCP service requires:

  1. Choosing a language (Node / Python / Rust, etc.)
  2. Implementing the MCP standard protocol (based on JSON-RPC)
  3. Writing a server.json declaration file locally
  4. Adding the service path to Pop's "Custom Service Directory"
  5. Refreshing services in Pop to load successfully

Example uses:

  • Connect your MySQL / Redis database
  • Integrate your company's internal APIs
  • Build automation script tools
  • Test tools for product development
  • Implement your private knowledge processing service

6. Service Status and Debugging

Pop provides powerful MCP debugging capabilities.

You can view:

  • Service startup logs
  • Tool registration status
  • AI tool invocation request records
  • JSON communication between AI and MCP services
  • Error stack traces
  • Tool time consumption analysis

This is very important for developing custom tools.

7. Frequently Asked Questions (FAQ)

Q: What to do if MCP service fails to start?

Possible reasons include:

  • Execution permission not granted
  • Node / Python environment missing
  • Port occupied
  • Incorrect service path

Checking logs can quickly locate the problem.

Q: Will AI have unlimited access to my files?

No.

By default, Pop:

  • Does not open write permissions
  • Does not open system root directories
  • Requires users to grant permissions manually

Q: Can custom MCP services be deployed remotely?

Yes, as long as the MCP server can be connected via local process or WebSocket.

Summary

MCP is Pop's extension engine, making AI no longer a pure chat model but having real tool usage capabilities.
Through MCP, you can let AI:

  • Read and write files
  • Scrape web pages
  • Access databases
  • Execute scripts
  • Integrate your custom APIs

It extends Pop's capabilities to almost unlimited heights and is a core tool for advanced and professional users.

Previous page
Model Settings
Next Page
Feedback