MCP Catalogs
Home

charlotte

by TickTockBent·142·Score 50

Token-efficient browser MCP server that provides structured web access to AI agents.

browser-automationweb-scrapingdeveloper-tools
21
Forks
13
Open issues
this month
Last commit
2d ago
Indexed

Overview

Charlotte is a specialized MCP server that dramatically reduces token consumption when AI agents interact with web pages. Instead of dumping the full accessibility tree (which can exceed 1M characters), Charlotte provides structured, typed representations with configurable detail levels. It maintains a persistent headless Chromium session and offers comprehensive navigation, observation, interaction, and development tools. Its benchmark results show 25-182x less data per page compared to Playwright MCP, resulting in significant cost savings for production workloads.

Try asking AI

After installing, here are 7 things you can ask your AI assistant:

you:AI agents browsing the web with minimal token consumption
you:Web automation that needs to avoid excessive API calls
you:Web development and accessibility testing for AI systems
you:Cost-effective web scraping for large language models
you:How does Charlotte reduce token consumption?
you:What browser does Charlotte use?
you:Can I use Charlotte with other MCP clients?

When to choose this

Choose Charlotte when token efficiency is critical and you need structured web access with controlled detail levels for AI agents.

When NOT to choose this

Don't choose Charlotte if you need raw HTML access or low-level DOM manipulation beyond what its structured representations provide.

Tools this server exposes

12 tools extracted from the README
  • navigate

    Navigate to a URL and get a minimal page summary

  • observe

    Observe the current page with configurable detail levels

  • find

    Find elements by type, text, or CSS selector

  • click

    Click on a specific element

  • type

    Type text into a text input or textarea

  • submit

    Submit a form by its ID

  • screenshot

    Take a screenshot of the current page

  • fill_form

    Fill multiple fields in a form at once

  • tab_open

    Open a new tab

  • dev_serve

    Start a local development server

  • dev_audit

    Run accessibility, performance, or SEO audits

  • evaluate

    Execute JavaScript in the page context

Comparable tools

playwright-mcpbrowser-mcpscraping-mcp

Installation

Installation

Charlotte is available on npm and Docker:

# Install with npm
npm install -g @ticktockbent/charlotte

# Or with Docker
docker pull ticktockbent/charlotte:alpine

Claude Desktop Configuration

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "charlotte": {
      "command": "npx",
      "args": ["@ticktockbent/charlotte"]
    }
  }
}

FAQ

How does Charlotte reduce token consumption?
Charlotte provides structured page representations with configurable detail levels instead of dumping the full accessibility tree. It returns minimal orientation data by default and only provides specific details when explicitly requested.
What browser does Charlotte use?
Charlotte maintains a persistent headless Chromium session as its rendering engine.
Can I use Charlotte with other MCP clients?
Yes, Charlotte supports multiple MCP clients including Claude Code, Cursor, Windsurf, VS Code with Copilot, Cline, and Amp.

Compare charlotte with

GitHub →

Last updated · Auto-generated from public README + GitHub signals.