linkedin-spider

A modern LinkedIn scraping library with built-in anti-detection, available as a Python library, CLI tool, and MCP server.

MCP Server · Python Library · CLI · Docker

Features

• Profile search with advanced filters (location, industry, company, connection degree)
• Post search by keywords with engagement metrics and comments
• Profile scraping with experience, education, skills, and contact details
• Company scraping with industry, size, specialties, and more
• Connection management — retrieve and send connection requests
• Conversations — list threads, read message history, and send messages
• Proxy support — route traffic through HTTP or SOCKS5 proxies
• Anti-detection — human-like behavior simulation, stealth mode, session persistence

Installation

pip install linkedin-spider        # library only
pip install linkedin-spider[cli]   # library + CLI
pip install linkedin-spider[mcp]   # library + MCP server
pip install linkedin-spider[all]   # everything

Authentication

linkedin-spider supports two authentication methods. Sessions are persisted in a Chrome profile so you typically only authenticate once.

LinkedIn Cookie (Recommended)

1. Log in to LinkedIn in your browser
2. Open DevTools (F12) → Application → Cookies → linkedin.com
3. Copy the li_at cookie value

scraper = LinkedinSpider(li_at_cookie="your_cookie_value")

Email & Password

scraper = LinkedinSpider(email="you@example.com", password="your_password")

For CLI and MCP, pass --cookie (or --email/--password) flags, or set the LINKEDIN_COOKIE (or LINKEDIN_EMAIL/LINKEDIN_PASSWORD) environment variables.

MCP Server

The MCP server exposes LinkedIn data to AI assistants like Claude.

Example prompts you can give Claude once connected:

Research the background of this candidate https://www.linkedin.com/in/johndoe

Find 10 product managers in San Francisco and summarize their experience

What has OpenAI been posting about recently? https://www.linkedin.com/company/openai

Show me my pending connection requests and summarize who they are

Send a message to John Doe saying "Thanks for connecting!"

It provides 12 tools:

Tool	Description
search_profiles	Search profiles with filters (location, industry, company, connections)
scrape_profile	Extract complete profile data from a URL
search_posts	Search posts by keywords with date filters
scrape_company	Get company details from a URL
scrape_incoming_connections	List pending connection requests received
scrape_outgoing_connections	List connection requests you've sent
send_connection_request	Send a connection request with optional note
scrape_conversations_list	List messaging conversations
scrape_conversation	Read messages from a conversation
send_message	Send a message in an existing or new conversation
get_session_status	Check if the browser session is active
reset_session	Close and reset the browser session

Start the server

# stdio (default — for Claude Desktop, Claude Code)
linkedin-spider-mcp serve --cookie your_li_at_cookie_value

# SSE or HTTP (for remote clients)
linkedin-spider-mcp serve --transport sse --host 127.0.0.1 --port 8000
linkedin-spider-mcp serve --transport http --host 0.0.0.0 --port 9000

Or configure via environment variables in a .env file:

LINKEDIN_COOKIE=your_li_at_cookie_value
HEADLESS=true
PROXY_URL=http://host:port          # optional

Claude Desktop

Add to your Claude Desktop config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json
• Linux: ~/.config/Claude/claude_desktop_config.json

With Docker (Recommended)

No local dependencies required — the Docker image includes Python, Chrome, and everything needed.

{
  "mcpServers": {
    "linkedin-spider": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "LINKEDIN_COOKIE=your_li_at_cookie_value",
        "-e",
        "HEADLESS=true",
        "-e",
        "TRANSPORT=stdio",
        "vertexcoverlabs/linkedin-mcp"
      ]
    }
  }
}

Without Docker

Requires pip install linkedin-spider[mcp] and Chrome installed locally.

{
  "mcpServers": {
    "linkedin-spider": {
      "command": "linkedin-spider-mcp",
      "args": ["serve", "--cookie", "your_li_at_cookie_value"]
    }
  }
}

Claude Code

With Docker

claude mcp add linkedin-spider -- \
  docker run --rm -i \
  -e LINKEDIN_COOKIE=your_li_at_cookie_value \
  -e HEADLESS=true \
  -e TRANSPORT=stdio \
  vertexcoverlabs/linkedin-mcp

Without Docker

claude mcp add linkedin-spider -- \
  linkedin-spider-mcp serve --cookie your_li_at_cookie_value

Or connect to a running SSE/HTTP server:

claude mcp add linkedin-spider --transport sse http://localhost:8000/sse

Troubleshooting

Authentication issues

• Cookie expired: LinkedIn cookies expire periodically. Grab a fresh li_at value from your browser.
• Email/password not working: LinkedIn may trigger a CAPTCHA or verification. Try cookie auth instead.
• Session reuse: Sessions are saved in a Chrome profile. If things break, delete the profile directory and re-authenticate.

Docker issues

• First-time pull is slow: The image is ~1.4GB. Pre-pull with docker pull vertexcoverlabs/linkedin-mcp before configuring Claude Desktop to avoid timeout.
• Port conflicts: If port 8080 is in use, map to a different host port: -p 9090:8080.

Browser / Chrome issues

• Chrome not found (non-Docker): Set CHROMEDRIVER_PATH in your .env or pass chromedriver_path in ScraperConfig.
• Page load timeouts: Increase page_load_timeout in ScraperConfig or use a faster proxy.
• Headless mode issues: Some LinkedIn pages behave differently in headless mode. Try headless=False for debugging.

Python Library

from linkedin_spider import LinkedinSpider, ScraperConfig

config = ScraperConfig(headless=True, page_load_timeout=30)
scraper = LinkedinSpider(
    li_at_cookie="your_cookie_value",
    config=config,
)

# Search profiles
results = scraper.search_profiles("software engineer", max_results=10)

# Search posts
posts = scraper.search_posts("artificial intelligence", max_results=10)

# Scrape a single profile
profile = scraper.scrape_profile("https://linkedin.com/in/someone")

# Scrape a company page
company = scraper.scrape_company("https://linkedin.com/company/openai")

# Connection requests
incoming = scraper.scrape_incoming_connections(max_results=20)
scraper.send_connection_request("https://linkedin.com/in/someone", note="Hi!")

# Conversations
threads = scraper.scrape_conversations_list(max_results=10)
messages = scraper.scrape_conversation_messages("John Doe")
scraper.send_message("Thanks for connecting!", participant_name="John Doe")

# Always clean up
scraper.close()

See the examples/ directory for more detailed usage.

Configuration

ScraperConfig accepts the following options:

Option	Default	Description
headless	False	Run browser without a visible window
stealth_mode	True	Inject anti-detection scripts
page_load_timeout	30	Page load timeout in seconds
implicit_wait	10	Implicit wait timeout in seconds
human_delay_range	(0.5, 2.0)	Random delay range between actions
proxy	None	Proxy URL (http:// or socks5://)
custom_user_agent	None	Override the default user agent

Command Line Interface

# Search profiles
linkedin-spider-cli search -q "product manager" -n 10 -o results.json

# Search posts
linkedin-spider-cli search-posts -k "artificial intelligence" -n 10 -o posts.json

# Scrape a profile
linkedin-spider-cli profile -u "https://linkedin.com/in/johndoe" -o profile.json

# Scrape a company
linkedin-spider-cli company -u "https://linkedin.com/company/openai" -o company.json

# List connection requests
linkedin-spider-cli connections -n 20 -o connections.json

Pass --cookie on first use (or set LINKEDIN_COOKIE env var). You can also use --email/--password instead. The session is saved and reused for subsequent commands.

Output defaults to stdout. Use -o to write JSON or CSV files.

Docker

A pre-built Docker image is published on Docker Hub:

docker pull vertexcoverlabs/linkedin-mcp

Or build locally:

docker build -t linkedin-spider .

Run with different transports:

# stdio
docker run --rm -i -e TRANSPORT=stdio --env-file .env vertexcoverlabs/linkedin-mcp

# SSE
docker run -p 8080:8080 -e TRANSPORT=sse --env-file .env vertexcoverlabs/linkedin-mcp

# HTTP
docker run -p 8080:8080 -e TRANSPORT=http --env-file .env vertexcoverlabs/linkedin-mcp

Running in the Cloud

LinkedIn blocks requests from known datacenter IP ranges. To run linkedin-spider on a cloud server (AWS, GCP, Azure, etc.), route browser traffic through a residential or mobile proxy using the PROXY_URL environment variable.

LINKEDIN_COOKIE=your_li_at_cookie_value
HEADLESS=true
PROXY_URL=http://user:pass@proxy-host:port

Both HTTP and SOCKS5 proxies are supported:

PROXY_URL=http://user:pass@proxy-host:port
PROXY_URL=socks5://user:pass@proxy-host:port

When using Docker, pass it as an environment variable:

docker run --rm -i \
  -e LINKEDIN_COOKIE=your_li_at_cookie_value \
  -e PROXY_URL=http://user:pass@proxy-host:port \
  -e HEADLESS=true \
  -e TRANSPORT=stdio \
  vertexcoverlabs/linkedin-mcp

When using the Python library, pass it via ScraperConfig:

config = ScraperConfig(headless=True, proxy="http://user:pass@proxy-host:port")

[!TIP] Residential proxies are recommended over datacenter proxies to avoid detection.

Development

git clone https://github.com/vertexcover-io/linkedin-spider
cd linkedin-spider
uv sync
cp .env.example .env   # add your credentials

make check    # lint + typecheck
make test     # run tests
make build    # build wheel

Disclaimer

This tool is for personal and educational use. Please respect LinkedIn's Terms of Service, use reasonable rate limits, and handle collected data responsibly.