Linkedin

LinkedIn MCP Skill

Browser-automated LinkedIn integration that scrapes profiles, searches jobs, sends messages, and analyzes companies through a persistent authenticated session.

When to Use This Skill

• Profile Research: Fetch detailed person profiles including experience, education, and skills
• Job Search: Search and retrieve job listings with filters
• Company Analysis: Get company profiles, employee counts, and recent posts
• People Search: Find professionals by role, location, company, or keywords
• Messaging: Send connection requests or messages to existing connections
• Inbox Management: Read conversations and search message history

When Not To Use

• For general web scraping or non-LinkedIn sites -- use the playwright or browser skills instead
• For job board aggregation across multiple platforms -- handle manually or use dedicated aggregators
• For LinkedIn Ads management or campaign creation -- not supported by this tool
• For bulk automated outreach or spam -- violates LinkedIn ToS; this skill is for targeted research
• For data that requires LinkedIn Premium or Sales Navigator -- session uses your account tier

Architecture

┌─────────────────────────────────┐
│  Claude Code / Skill Invocation │
└──────────────┬──────────────────┘
               │ MCP Protocol (stdio)
               ▼
┌─────────────────────────────────┐
│  LinkedIn Scraper MCP Server    │
│  (uvx / Patchright)            │
└──────────────┬──────────────────┘
               │ Browser Automation
               ▼
┌─────────────────────────────────┐
│  Chromium (Patchright-managed)  │
│  Persistent profile at          │
│  ~/.linkedin-mcp/profile/       │
└──────────────┬──────────────────┘
               │ HTTPS
               ▼
┌─────────────────────────────────┐
│  LinkedIn.com                   │
└─────────────────────────────────┘

Authentication

No API key required. The server uses browser-based login with a persistent Chromium profile stored at ~/.linkedin-mcp/profile/. On first use, the browser opens a login page; after authenticating once, the session persists across restarts.

A private Patchright Chromium is a scoped exception, not a policy breach. This skill drives its own Patchright-managed browser instead of the browsercontainer GPU sidecar because LinkedIn ties a session to a persistent per-site profile — cookies, device fingerprint, and login state that must survive restarts, or every call re-triggers a login prompt or CAPTCHA. The shared sidecar's session model is not built for that kind of long-lived, single-site auth state shared across unrelated tasks, so it is not a substitute here (see the notebooklm skill's equivalent exception for its own OAuth browser). This is a per-site auth-persistence requirement, not general-purpose scraping or automation, so the container's sidecar-only browser policy does not apply to it. The ToS risk stands regardless of browser choice: LinkedIn prohibits automated scraping and bulk messaging, so keep usage to targeted, low-volume research and account-holder-authorised messaging (see When Not To Use below) — routing through the sidecar would not remove that risk, only the profile-persistence problem is specific to the local browser.

Tools

Examples

# Fetch a person's profile
get_person_profile({
    "url": "https://www.linkedin.com/in/satyanadella/"
})

# Search for jobs
search_jobs({
    "keywords": "machine learning engineer",
    "location": "San Francisco Bay Area",
    "job_type": "full-time"
})

# Search for people at a company
search_people({
    "keywords": "VP Engineering",
    "company": "Anthropic"
})

# Get company profile and recent posts
get_company_profile({
    "url": "https://www.linkedin.com/company/anthropic/"
})

# Send a connection request
connect_with_person({
    "profile_url": "https://www.linkedin.com/in/example/",
    "message": "Would love to connect regarding our shared interest in AI safety."
})

Environment Variables

Setup

# Install via uvx (recommended)
uvx linkedin-scraper-mcp

# Or add to MCP config
# The server will prompt for browser login on first run

Troubleshooting

Login Session Expired:

# Remove persistent profile and re-authenticate
rm -rf ~/.linkedin-mcp/profile/
# Restart the MCP server -- browser will open login page

Browser Not Found:

# Install Patchright browsers
uvx patchright install chromium

# Or specify Chrome path explicitly
export CHROME_PATH="$(which chromium-browser || which chromium)"

Rate Limiting / CAPTCHA:

• LinkedIn may throttle automated browsing; reduce request frequency
• If CAPTCHA appears, solve it manually in the browser window
• The persistent session reduces CAPTCHA frequency significantly

Timeout Errors:

# Increase timeout for slow connections
export LINKEDIN_TIMEOUT=60000

Integration with Other Skills

Combine with:

• perplexity-research: Cross-reference LinkedIn profiles with web research
• gemini-url-context: Expand company blog posts found via LinkedIn
• report-builder: Generate hiring pipeline or competitive analysis reports