Skip to content

Secure-tree/AI-claude-playwright-Linkedin--MCP-sever-Automation

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

🔗 kaushik-linkedin-mcp

A personal LinkedIn MCP server — built from scratch. Not a fork. Not a clone. Written by Kaushik Muthukumaran.

Python FastMCP Playwright License Security

This MCP server connects Claude AI directly to LinkedIn using browser automation. It gives Claude 8 tools to read profiles, search people and jobs, get company data, publish posts, edit your profile, and manage sessions — all from a conversation with Claude.


📖 Table of Contents

  1. What You Can Do With This
  2. How to Clone and Run It
  3. Project Structure
  4. How It Works
  5. Tools Reference
  6. For People Cloning This Repo
  7. Security
  8. Troubleshooting
  9. Author

🚀 What You Can Do With This

Once connected to Claude, you can have natural conversations that do real LinkedIn actions.


✏️ Edit Your LinkedIn Profile By Just Telling Claude

Method 1 — Tell Claude your details in chat:

"Change my headline to: Security Engineer | Zero Trust | CrowdStrike | 14K+ Endpoints" "Add a new job: Security Engineer at MasTec, promoted after 1 year" "Update my Sankara Nethralaya role to focus on vulnerability assessments and OSINT"

Claude opens Chrome, finds the right field, types your content, and saves it.

Method 2 — Upload your resume and ask Claude to sync it:

"Here's my updated resume — update my LinkedIn to match it"

Claude reads every section, compares it to your LinkedIn, and edits each section one by one. Supports: .pdf, .docx, .txt

Method 3 — Targeted corrections mid-conversation:

"Split my MasTec role into two entries at August 2025" "Add OSINT and CyberArk to my skills" "Move Sankara Nethralaya from Part-time to Full-time"


👤 Profile Intelligence

"Pull Kaushik's full LinkedIn profile including experience and skills" "Show me their recent posts and engagement"

Reads: name, headline, location, connections, about, full experience, education, skills, posts, contact info.


🔍 People Search

"Find Security Engineers at CrowdStrike in Pennsylvania" "Search for CISOs in the healthcare industry"

Returns: name, headline, location, username, profile URL. Max 100 results per call.


💼 Job Search & Research

"Find Senior Security Engineer jobs in Pennsylvania posted this week" "Show me remote Zero Trust architect roles — mid-senior level only" "What's the salary range for this CrowdStrike job?"

Full LinkedIn job search: date posted, experience level, job type, work arrangement, Easy Apply.


🏢 Company Research

"What does CrowdStrike's LinkedIn page say about them?" "Show me MasTec's recent posts and engagement numbers" "What roles is Microsoft hiring for right now?"

Returns: about, industry, size, headquarters, recent posts with reactions, open roles.


📢 Post Publishing

"Publish my vulnerability management post to LinkedIn" "Post this Zero Trust insight with these hashtags"

Navigates to feed, opens composer, injects text securely, clicks Post, confirms success.


🤖 Combined Power Workflows

"Search for Senior Security Engineer jobs at CrowdStrike, get the top 3 JDs, and tell me exactly how my resume matches each one"

"Find Security Engineers at my target companies and draft a personalised connection message for each one based on their background"

"Write and publish a post about today's Rapid7 CVE findings — under 200 words with 5 hashtags"


🖥️ How to Clone and Run It

Step 1 — Install Homebrew

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
echo 'eval "$(/opt/homebrew/bin/brew shellenv zsh)"' >> ~/.zprofile
eval "$(/opt/homebrew/bin/brew shellenv zsh)"

Step 2 — Install Node.js and Git

brew install node git

Step 3 — Install uv

curl -LsSf https://astral.sh/uv/install.sh | sh
source $HOME/.local/bin/env

⚠️ Save the path from which uv — you'll need it for Claude Desktop config.

Step 4 — Clone and Install

git clone https://github.com/Secure-tree/kaushik-linkedin-mcp
cd kaushik-linkedin-mcp
uv sync
uv run playwright install chromium

Step 5 — Authenticate with LinkedIn

uv run kaushik-linkedin-mcp --login

A Chrome window opens. Log in to your LinkedIn. Session saves automatically.

Step 6 — Verify Session

uv run kaushik-linkedin-mcp --status
# Status: Valid  ✅

Step 7 — Connect to Claude Desktop

Open: ~/Library/Application Support/Claude/claude_desktop_config.json

Merge this in:

{
  "mcpServers": {
    "linkedin": {
      "command": "/Users/YOUR_USERNAME/.local/bin/uv",
      "args": [
        "--directory",
        "/Users/YOUR_USERNAME/kaushik-linkedin-mcp",
        "run",
        "kaushik-linkedin-mcp"
      ]
    }
  }
}

⚠️ Use the full path to uv from which uv — Claude Desktop does not inherit your terminal PATH.

Restart Claude:

pkill -f Claude && sleep 2 && open /Applications/Claude.app

🏗️ Project Structure

kaushik-linkedin-mcp/
├── pyproject.toml                      # Project config, CLI entry point, dependencies
├── .python-version                     # Pins Python 3.12
├── .env.example                        # Config template — copy to .env
├── .gitignore                          # Protects session data and .env
├── LICENSE                             # MIT
└── src/
    └── kaushik_linkedin_mcp/
        ├── __init__.py
        ├── server.py                   # FastMCP server — all 8 tools
        ├── browser.py                  # Playwright session manager
        └── tools/
            ├── profile.py              # get_person_profile
            ├── jobs.py                 # search_jobs + get_job_details
            ├── posts.py                # publish_linkedin_post
            └── search.py              # search_people + get_company_profile

⚙️ How It Works

You (in Claude chat)
        │
        │  "Search for Security Engineer jobs in Pennsylvania"
        ▼
Claude Desktop
        │  stdio (local process)
        ▼
kaushik-linkedin-mcp (Python / FastMCP)
        │  Playwright browser automation
        ▼
Chromium (headless, on your Mac)
        │  HTTPS — your authenticated session
        ▼
LinkedIn.com → returns structured JSON back to Claude

Note for cloners: Session saves to ~/.kaushik-linkedin-mcp/session/storage.json on YOUR machine. The author has zero access to your account.


🧰 Tools Reference

Tool Parameters
get_person_profile linkedin_username, sections (main/experience/education/skills/posts/contact)
search_linkedin_people keywords, location, max_results (1–100)
search_linkedin_jobs keywords, location, date_posted, experience_level, job_type, work_type, easy_apply, max_pages
get_linkedin_job_details job_id
get_linkedin_company company_name, sections (about/posts/jobs)
publish_linkedin_post text
check_session_status
close_browser

👥 For People Cloning This Repo

When you clone and run this, you log into your own LinkedIn — not the author's.

Thing Change Required
--login step Log in with YOUR LinkedIn credentials
Claude Desktop config Update path to YOUR clone location and uv path
.env Optional — defaults work for most setups
pyproject.toml author Leave as-is — this is the author's name

Quick start:

git clone https://github.com/Secure-tree/kaushik-linkedin-mcp
cd kaushik-linkedin-mcp
uv sync
uv run playwright install chromium
uv run kaushik-linkedin-mcp --login
uv run kaushik-linkedin-mcp --status

🔐 Security

Check Status
No hardcoded credentials
No JS template literal injection ✅ User values passed via CDP args
Input validation ✅ max_results clamped 1–100
Session stored outside repo ~/.kaushik-linkedin-mcp/
.env never committed ✅ In .gitignore
No shell/subprocess calls
No 3rd party API calls ✅ Only linkedin.com
Local stdio communication ✅ No network exposure

🔧 Troubleshooting

No LinkedIn tools in Claude → Use full uv path in config: which uv/Users/YOU/.local/bin/uv

Status: Expired/Invaliduv run kaushik-linkedin-mcp --logout && uv run kaushik-linkedin-mcp --login

npm install fails → This is Python. Use uv sync not npm install.

Config has two {} blocks → Merge into one root object: { "preferences": {...}, "mcpServers": {...} }


⚡ Quick Reference

# Setup
git clone https://github.com/Secure-tree/kaushik-linkedin-mcp
cd kaushik-linkedin-mcp && uv sync
uv run playwright install chromium
uv run kaushik-linkedin-mcp --login
uv run kaushik-linkedin-mcp --status

# Restart Claude after config changes
pkill -f Claude && sleep 2 && open /Applications/Claude.app

# Re-authenticate
uv run kaushik-linkedin-mcp --logout && uv run kaushik-linkedin-mcp --login

👤 Author

Kaushik Muthukumaran Security Engineer | AI + Cybersecurity Researcher

  • 🔗 LinkedIn
  • 📧 kaushikrmk01@gmail.com
  • 🏅 Ms in Cybersecurity
  • 🏅 CompTIA Network+ Certified
  • 🔬 Research: AI-Driven Threat Detection · Zero Trust · ML Risk Scoring and Remediation

📄 License

MIT License — see LICENSE for details.


Built with Claude AI (Anthropic) · FastMCP · Playwright · Python 3.12

About

A personal LinkedIn MCP server built from scratch — connects Claude AI to LinkedIn for profile editing, job search, people search, company research, and post publishing

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages