comet-mcp
An MCP server that connects Claude to Perplexity Comet for agentic web browsing and deep research via an isolated browser instance. It enables Claude to delegate complex navigation, login flows, and real-time task monitoring to Perplexity's specialized research intelligence.
README
comet-mcp
<a href="https://glama.ai/mcp/servers/@hanzili/comet-mcp"> <img width="380" height="200" src="https://glama.ai/mcp/servers/@hanzili/comet-mcp/badge" /> </a>
Give Claude Code a browser that thinks.
An MCP server that connects Claude Code to Perplexity Comet - enabling agentic web browsing, deep research, and real-time task monitoring.

What's New in v3.0
- Isolated MCP Instance - Uses a dedicated Comet browser (port 9223) - your personal tabs are never touched
- Full Research Text - Fixed truncation issue, now returns complete research responses
- Reliable Startup - Auto-retry logic, kills stale processes, extended timeouts
- Login Detection - First-time setup prompts for Perplexity sign-in
- Visual Indicator - MCP Comet shows "[MCP]" badge so you know which browser is which
- Research Folders - Save and organize research into folders
- Library Search - Search your existing Perplexity research history
Why?
Existing web tools for Claude Code fall into two categories, both with limitations:
1. Search APIs (Tavily, Perplexity API, WebFetch)
Return static text. No interaction, no login, no dynamic content. Great for quick lookups, but can't navigate complex sites or fill forms.
2. Browser Automation (browser-use, Puppeteer MCP, Playwright MCP)
Can interact with pages, but use a one-agent-do-all approach: the same reasoning model that's writing your code is also deciding where to click, what to type, and how to navigate. This overwhelms the context window and fragments focus.
3. Comet MCP: Multi-Agent Delegation
Comet MCP takes a different approach. Instead of Claude controlling a browser directly, it delegates to Perplexity Comet - an AI purpose-built for web research and browsing.
- Claude stays focused on your coding task
- Comet handles the browsing: navigation, login walls, dynamic content, deep research
- Result: Claude's coding intelligence + Perplexity's web intelligence, working together
Installation (3 Steps)
Step 1: Install Comet Browser
Download: https://www.perplexity.ai/comet
Step 2: Add to Claude Code
Add to ~/.claude.json:
{
"mcpServers": {
"comet": {
"command": "npx",
"args": ["-y", "comet-mcp"]
}
}
}
Step 3: macOS Setup (One-time)
Run the setup script to create an isolated MCP browser:
curl -sL https://raw.githubusercontent.com/hanzili/comet-mcp/main/scripts/setup-macos.sh | bash
<details> <summary>Or run manually</summary>
MCP_APP="$HOME/.comet-mcp/Comet-MCP.app"
mkdir -p "$HOME/.comet-mcp"
cp -R "/Applications/Comet.app" "$MCP_APP"
/usr/libexec/PlistBuddy -c "Set :CFBundleIdentifier ai.perplexity.comet.mcp" "$MCP_APP/Contents/Info.plist"
codesign --force --deep --sign - "$MCP_APP"
</details>
First Use
- Restart Claude Code
- Say: "Connect to Comet"
- Login to Perplexity in the MCP window (first time only)
- Done!
Try It
You: "Use Comet to research the top AI frameworks in 2025"
Claude: [delegates to Comet, monitors progress, returns results]
You: "Log into my GitHub and check my notifications"
Claude: [Comet handles the login flow and navigation]
You: "Save this research to my 'AI Projects' folder"
Claude: [uses comet_folders to organize research]
Tools
| Tool | Description |
|---|---|
comet_connect |
Connect to MCP-dedicated Comet (auto-starts, isolated from personal browser) |
comet_ask |
Send a task and wait for response |
comet_poll |
Check progress on long-running tasks |
comet_stop |
Stop current task |
comet_screenshot |
Capture current page |
comet_mode |
Switch modes: search, research, labs, learn |
comet_folders |
List, create, or save to research folders |
comet_library |
Search your existing Perplexity research |
How It Works
Claude Code → MCP Server → CDP (port 9223) → MCP Comet → Perplexity AI
(reasoning) (bridge) (isolated) (web browsing)
- Port 9223: MCP uses its own port, never touches your personal Comet (9222)
- Separate Profile: Data stored in
~/.comet-mcp, completely isolated - Visual Badge: MCP Comet shows "[MCP]" indicator so you know which is which
Claude sends high-level goals ("research X", "log into Y"). Comet figures out the clicks, scrolls, and searches. Results flow back to Claude.
Requirements
- Node.js 18+
- Perplexity Comet Browser
- Claude Code (or any MCP client)
- Supported platforms: macOS, Windows, WSL2
Windows & WSL Support
Native Windows
Works out of the box. Comet MCP auto-detects Windows and launches Comet from its default install location.
WSL2 (Windows Subsystem for Linux)
WSL2 requires mirrored networking to connect to Comet running on Windows:
-
Enable mirrored networking (one-time setup):
# Create/edit %USERPROFILE%\.wslconfig (Windows side) [wsl2] networkingMode=mirrored -
Restart WSL:
wsl --shutdown # Then reopen your WSL terminal -
That's it! Comet MCP auto-detects WSL and uses PowerShell to communicate with Windows.
If mirrored networking isn't available, you'll see a helpful error message with setup instructions.
Custom Comet Path
If Comet is installed in a non-standard location:
{
"mcpServers": {
"comet-bridge": {
"command": "npx",
"args": ["-y", "comet-mcp"],
"env": {
"COMET_PATH": "/path/to/your/Comet"
}
}
}
}
Troubleshooting
"Cannot connect to Comet"
- macOS: Ensure Comet is installed at
/Applications/Comet.appand you've run the one-time setup to create~/.comet-mcp/Comet-MCP.app - Windows: Comet should be in
%LOCALAPPDATA%\Perplexity\Comet\Application\ - MCP uses port 9223 (not 9222) - check if available
"MCP Comet closes my personal Comet" (macOS)
- You need to create the separate app bundle first (see "macOS: Create Isolated App Bundle" above)
- Without this, Electron's single-instance lock will replace your personal browser
"Not logged in" message
- Log into Perplexity in the MCP Comet browser window (the one with [MCP] badge)
- Then call
comet_connectagain
"WSL cannot connect to Windows localhost"
- Enable mirrored networking (see WSL section above)
- Or run Claude Code from Windows PowerShell instead of WSL
"Tools not showing in Claude"
- Restart Claude Code after config changes
"Research text is truncated"
- Upgrade to v3.0.0 - this issue is fixed
- Use
comet_pollfor long research to get full results
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。