Skip to main content
The MCP-B extension includes four specialized AI agents, each designed for specific types of tasks. Choosing the right agent helps you work more effectively with web automation and MCP integration.

Available Agents

Userscript Engineer

Build custom scripts to enhance any website

WebMCP Server

Turn websites into AI-accessible tools

Browsing Agent

Navigate and inspect web pages

Chat Companion

Ask questions without automation

Switching Between Agents

You can switch agents at any time in the MCP-B extension:
1

Open the extension

Click the MCP-B icon in your browser toolbar
2

Access agent selector

Look for the agent dropdown menu at the top of the chat interface
3

Select your agent

Choose the agent that matches your current task
4

Start chatting

Describe what you want to accomplish
Your conversation history is preserved when switching agents, but each agent has different capabilities and tool access.

Userscript Engineer

Best for: Customizing websites, adding features, automating repetitive tasks

What it Does

The Userscript Engineer helps you create custom scripts that run on websites to modify their appearance or add new functionality. Think of it as having a personal developer who can customize any website for you.

Example Uses

Add Features

Add a dark mode toggle, custom shortcuts, or missing features to any site

Remove Clutter

Hide ads, banners, or distracting elements from your favorite sites

Automate Tasks

Auto-fill forms, extract data, or streamline repetitive workflows

Customize Style

Change fonts, colors, layouts to match your preferences

Sample Requests

“Create a dark mode toggle button for GitHub that changes the background to dark and text to light colors”
“Remove the ‘Who to follow’ and ‘Trends’ sections from Twitter’s sidebar”
“Create a script that auto-fills my shipping address on checkout pages”
Userscripts modify how websites look and behave in your browser. They only affect your view - they don’t change the website for anyone else.

WebMCP Server Agent

Best for: Making websites work with AI agents, creating structured tools from web functionality

What it Does

The WebMCP Server agent builds special userscripts that expose website features as “tools” that AI assistants can use. Instead of just modifying a page, these scripts make website functionality programmatically accessible.

Example Uses

Search Integration

Let AI search a site and retrieve results automatically

Data Extraction

Pull structured data from websites for AI analysis

Action Automation

Enable AI to click buttons, submit forms, or navigate sites

Workflow Chains

Chain multiple website actions together for complex tasks

How it’s Different from Userscripts

Userscript EngineerWebMCP Server
Changes what you seeMakes features accessible to AI
Adds buttons/stylingAdds invisible tools
For human interactionFor AI agent interaction
Visual modificationsProgrammatic access

Sample Requests

“Create MCP tools so AI can search Amazon products and get price information”
“Build tools that let AI extract profile data from LinkedIn search results”
“Create MCP tools so AI can post tweets on my behalf when I ask”
After the WebMCP Server agent creates tools, they automatically appear in your agent’s available tools list. You can then ask any agent to use those tools.
Web Standard APIs: The WebMCP Server agent uses navigator.modelContext.registerTool() - the W3C Web Model Context API standard. All userscripts are moving to this standard API, which accepts Zod schemas (preferred) directly in the inputSchema object (e.g., { param: z.string() }) as well as JSON Schema for input validation.

Browsing Agent

Best for: Investigating pages, gathering information, navigating websites

What it Does

The Browsing Agent helps you explore websites, extract information, and navigate between pages. It’s less about building scripts and more about using the browser effectively.

Example Uses

Research

Extract content from multiple pages for analysis

Monitoring

Check page status or content changes

Navigation

Help navigate complex websites

Screenshots

Capture and analyze page visuals

Sample Requests

“Go to these three product pages and compare their features and prices”
“Check this page every hour and let me know if the price changes”

Chat Companion

Best for: Asking questions, brainstorming, explanations without automation

What it Does

The Chat Companion is a general conversational assistant with minimal automation. It’s designed for questions, explanations, and planning - not for browser actions.

When to Use It

Choose Chat Companion when you:
  • Want to ask questions about web development or MCP
  • Need explanations without triggering browser actions
  • Want to plan before taking action
  • Prefer conversation over automation

Sample Requests

“What’s the difference between an MCP server and a userscript?”
“Help me plan out what features I should add to my website”
“Explain how the Model Context Protocol works”
If you need browser actions while chatting, the Chat Companion will suggest switching to a more appropriate agent.

Choosing the Right Agent

Use this guide to pick the best agent for your task:

Quick Comparison

  • Userscript Engineer
  • WebMCP Server
  • Browsing Agent
  • Chat Companion
Purpose: Customize and enhance websitesCreates: Scripts that modify pagesOutput: Visual changes you can seeBest for: Adding features, removing elements, stylingTool Access: Full access to all extension tools

Working with Agents

Tips for Success

Instead of “make this page better,” try “add a dark mode toggle button that changes the background to #1a1a1a”
Tell the agent which website you want to work with: “Go to example.com and…”
For complex scripts, build and test one feature at a time
If a task requires different capabilities, don’t hesitate to switch agents mid-conversation

Common Workflows

1

Userscript workflow

Request feature → Agent inspects page → Writes script → Tests on page → You approve → Script is saved
2

WebMCP Server workflow

Request tools → Agent analyzes site → Builds tools one by one → Tests each tool → Tools become available → You can use them
3

Browsing workflow

Ask question → Agent navigates to page → Extracts information → Provides answer with screenshots

Managing Your Scripts

Once an agent creates a userscript or WebMCP server, you can manage it through the extension:
  • View: See all your scripts in the extension’s userscript manager
  • Edit: Request modifications by starting a new conversation
  • Delete: Remove scripts you no longer need
  • Export: Download scripts to share or back up
  • Import: Upload scripts from others or from backups

Managing Userscripts

Learn how to download, upload, and manage your userscripts

Troubleshooting

The page structure may have changed. Ask the agent to inspect the page again and update the selectors.
Selectors may not be stable. Ask the agent to use more reliable selectors like IDs or data attributes instead of generated class names.
Wait a few seconds after the script runs. If they still don’t appear, check the browser console for errors and report them to the agent.
You can switch agents at any time. The new agent will have access to your conversation history.

Resources