linkedin-mcp

What is this?

Nano LinkedIn MCP server for LLM agents & Cursor
A developer tool to automate LinkedIn workflows using agents. You can scrape profiles, companies, inbox, threads, and send messages via LinkedIn, all through a local MCP server. Built for pipeline automation and easy integration with other tools.

---

Features

• Scrape LinkedIn profiles (name, experience, etc.)
• Scrape company pages (industry, size, etc.)
• Scrape inbox threads and messages
• Send messages to users or threads
• Easily extendable for new LinkedIn actions

---

Quickstart

1. Install dependencies

You'll need Python 3.8+.

This project uses the Phantombuster service for LinkedIn automation. You need an account and API key from Phantombuster to use all features.

pip install python-dotenv requests pydantic

2. Set up environment variables

Create a .env file in the root:

PHANTOMBUSTER_API_KEY=your_phantombuster_key
LINKEDIN_COOKIE_LI=your_linkedin_cookie
LINKEDIN_BROWSER_AGENT=your_browser_user_agent

How to get these values:

• LINKEDIN_BROWSER_AGENT: This is your browser user agent. Open Google and search for "my browser user agent", then copy the value.
• LINKEDIN_COOKIE_LI: This is your LinkedIn cookie. In Chrome: open LinkedIn, right-click and choose "Inspect", go to the "Application" tab, select "Cookies" > "https://linkedin.com", find "li_at" and copy its value.

3. Run the server

python mcp_server/linkedin-server.py

4. Example: Call from an agent

You can connect the MCP server to your Agent

async with MCPServerStdio(
    params={
        "command": "python",
        "args": [
            "{pathtoproject}/linkedin-mcp/mcp_server/linkedin-server.py"
        ],
        "env": {
            "PYTHONPATH": "{pathtoproject}/linkedin-mcp",
            "PHANTOMBUSTER_API_KEY": phantombuster_api_key or "",
            "LINKEDIN_COOKIE_LI": linkedin_cookie_li or "",
            "LINKEDIN_BROWSER_AGENT": linkedin_browser_agent or ""
        },
        "description": "LinkedIn MCP server: scrap profile,. "
    }
) as server:
    await server.connect()

    tools = await server.list_tools()

    agent=Agent(
        name="Assistant",
        instructions="Use the tools to achieve the task",
        mcp_servers=[server]
    )

5. Add to Cursor

• Open Cursor
• Go to MCP settings
• Add a new server with:

"linkedin-server": {
    "command": "python",
    "args": [
        "{pathtoproject}/linkedin-server.py"
    ],
    "env": {
        "PYTHONPATH": "{pathtoproject}",
        "PHANTOMBUSTER_API_KEY": "",
        "LINKEDIN_COOKIE_LI": "",
        "LINKEDIN_BROWSER_AGENT": ""
    },
    "description": "LinkedIn MCP server: scrap profile,. "
}

---

Roadmap

• Scrape user posts
• Collect more detailed company info
• Support other providers (e.g., browser automation, not just Phantombuster)
• Integrate with LinkedIn Sales Navigator
• Add memory (agent context, history)
• Add cache for requests/results
• Add limits (rate limiting, quotas)

⚠️ Important: LinkedIn does not like automation. Always respect daily limits to avoid account restrictions:

Free LinkedIn:

• 100 messages per day
• 100 profiles per day
• 20 connection requests per day

Premium LinkedIn:

• 150 messages per day
• 200 profiles per day
• 40 connection requests per day

---

About the Author

Maksim K. — Building Radr: Cursor for Pipeline. Ex-Product Lead (YC, TeachStars). CEO at Radr. Contact: max@radr.xyz