LinkedIn MCP Server
devag7/linkedin-mcp
36 ferramentas LinkedIn para Claude e IA — profiles, jobs, messaging, network
Ver no GitHubServidor MCP TypeScript que fornece acesso programático completo ao LinkedIn com 36 ferramentas: perfis, mensagens, vagas, empresas e feed. Funciona com Claude, Cursor e qualquer cliente MCP compatível, sem dependências externas e pronto para deploy remoto.
Instalação
npx convez-ai devag7/linkedin-mcp1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Automação de pesquisa de perfis e prospecção B2B
- Integração de dados LinkedIn com assistentes IA para análise de redes
- Automação de mensagens e gerenciamento de contatos em larga escala
Veredito

MCP server bem-engineered com autenticação segura, rate limiting e cache. Limitação importante: Cloudflare bot-management bloqueia a API em muitas redes — validar compatibilidade antes de usar em produção.
README
A tradução para o português deste README ainda está na fila.
🔗 LinkedIn MCP
A comprehensive LinkedIn MCP server for AI assistants
Give Claude, Cursor, and any MCP-compatible AI assistant access to LinkedIn — profiles, messaging, jobs, companies, and more.
npx -y linkedin-mcp-tools --login # One-time setup
Quick Start · 36 Tools · Setup Guide · Authentication · Compare
✨ Why LinkedIn MCP?
| Feature | Description | |
|---|---|---|
| 🚀 | Remote-First | Zero install — add a URL to your Claude config and go. No Python, no Docker required. |
| 🔧 | 36 Tools | Profiles, messaging, jobs, companies, network, feed, and more. |
| 🔒 | Secure | Cookie + CSRF authentication, rate limiting, configurable CORS. |
| ⚡ | Lightweight | ~65KB bundle. TypeScript, official MCP SDK. |
| 🏗️ | Well-Engineered | LRU cache, token-bucket rate limiting, exponential backoff, structured errors. |
⚠️ Important Limitation: LinkedIn's Voyager API is protected by Cloudflare bot-management. On many networks/IPs, Cloudflare returns 302 redirect loops that cannot be resolved with plain HTTP — even with a valid cookie and full cookie jar. If you experience
"Too many redirects"errors, this is a Cloudflare block, not a bug. See Known Limitations for details.
🚀 Quick Start
Method 1: Remote URL (Recommended)
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"linkedin": {
"url": "https://your-deployed-url.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN"
}
}
}
}
Method 2: npx (No Install)
First, save your credentials once:
npx -y linkedin-mcp-tools --login
Then add to Claude config (no env vars needed!):
{
"mcpServers": {
"linkedin": {
"command": "npx",
"args": ["-y", "linkedin-mcp-tools"]
}
}
}
Method 3: Clone & Run
git clone https://github.com/devag7/linkedin-mcp.git
cd linkedin-mcp
npm install
npm run dev
🛠 Available Tools
👤 Profile Tools
| Tool | Description |
|---|---|
get_profile | Get any LinkedIn profile with full details (experience, education, skills) |
get_my_profile | Get the authenticated user's own profile |
get_profile_skills | Get detailed skills with endorsement counts |
get_profile_recommendations | Get recommendations given and received |
get_profile_activity | Get a user's recent posts and activity |
get_sidebar_profiles | Get "People also viewed" profile suggestions |
search_people | Search people with advanced filters (location, company, title) |
💬 Messaging Tools
| Tool | Description |
|---|---|
get_inbox | List inbox conversations with read/unread filters |
get_conversation | Read a specific conversation thread |
search_conversations | Search messages by keyword |
send_message | Send a message (supports multiline, rich text) |
reply_to_thread | Reply to an existing conversation by thread ID |
mark_conversation_read | Mark a conversation as read |
🏢 Company Tools
| Tool | Description |
|---|---|
get_company | Get company profile, description, and details |
get_company_posts | Get recent posts from a company's feed |
get_company_employees | List employees with filters |
search_companies | Search companies by keyword and industry |
get_company_jobs | Get open job positions at a company |
💼 Job Tools
| Tool | Description |
|---|---|
search_jobs | Search jobs with keyword, location, and experience filters |
get_job_details | Get detailed information about a job posting |
get_saved_jobs | Get the user's saved/bookmarked jobs |
get_job_applicants | Get applicant information (recruiter accounts) |
🤝 Network Tools
| Tool | Description |
|---|---|
connect_with_person | Send a connection request with optional personalized note |
get_connections | List 1st-degree connections, sorted by recently added |
get_pending_invitations | View sent and received connection invitations |
withdraw_invitation | Cancel a sent connection invitation |
accept_invitation | Accept a received connection invitation |
get_network_stats | Get network growth metrics and connection statistics |
📰 Feed & Content Tools
| Tool | Description |
|---|---|
get_feed | Get posts from the home feed |
create_post | Create a new text post |
react_to_post | React to a post (like, celebrate, support, etc.) |
comment_on_post | Add a comment to a post |
search_posts | Search posts by keyword or hashtag |
🔧 Utility Tools
| Tool | Description |
|---|---|
whoami | Get server info, auth status, and capabilities |
health_check | Check server health and LinkedIn connectivity |
get_notifications | Get recent LinkedIn notifications |
🔐 Authentication
One-Time Setup (Recommended)
Run the interactive login once — credentials are saved and reused automatically:
npx linkedin-mcp-tools --login
This guides you through pasting your LinkedIn cookie or OAuth token, and saves it to ~/.linkedin-mcp/credentials.json. You never need to set env vars again.
# Check your current auth status
npx linkedin-mcp-tools --status
# Clear saved credentials
npx linkedin-mcp-tools --logout
Cookie Auth (Works with All 36 Tools)
- Log in to LinkedIn in your browser
- Open DevTools (F12) → Application → Cookies →
linkedin.com - Copy the
li_atcookie value - Copy the
JSESSIONIDcookie value (required — used as CSRF token) - Run
linkedin-mcp --loginand paste when prompted
Note: Both
li_atandJSESSIONIDare required. The JSESSIONID provides CSRF protection that LinkedIn's API requires for authenticated requests.
OAuth 2.0 (Experimental)
OAuth support is experimental and intended for future official REST API tools. Currently, all 36 tools use LinkedIn's internal Voyager API which requires cookie auth.
- Create an app at LinkedIn Developer Portal
- Get your access token
- Run
linkedin-mcp --loginand select OAuth option
Environment Variables (Alternative)
For CI/Docker, you can use env vars instead (they override saved credentials):
export LINKEDIN_COOKIE="your_li_at_cookie_value"
export LINKEDIN_CSRF_TOKEN="your_jsessionid_value" # Required for cookie auth
# or
export LINKEDIN_ACCESS_TOKEN="your_oauth_token"
⚡ Comparison with Alternatives
| Feature | LinkedIn MCP | stickerdaniel/linkedin-mcp-server |
|---|---|---|
| Architecture | API-based (Voyager) | Browser-based (Patchright + Chromium) |
| Cloudflare bypass | ⚠️ Blocked on many IPs — Cloudflare 302 loops | ✅ Real browser passes Cloudflare |
| Transport | Remote HTTP + stdio | stdio + experimental HTTP |
| Install | npx linkedin-mcp-tools | Python + uv + Chromium (~500MB) |
| Language | TypeScript (official MCP SDK) | Python (fastmcp — 3rd party) |
| Tools | 36 (when Cloudflare allows) | 17 |
| Auth Methods | Cookie + CSRF (li_at + JSESSIONID) | Browser login |
| Rate Limiting | ✅ Built-in token bucket | ❌ None |
| Response Caching | ✅ LRU with TTL | ❌ None |
| Retry Logic | ✅ Exponential backoff | ❌ None |
| Bundle Size | ~65KB | ~500MB+ (includes Chromium) |
Trade-off: This project is lightweight and well-engineered but relies on plain HTTP to LinkedIn's Voyager API. The browser-based alternative is heavier but bypasses Cloudflare because it uses a real browser. Choose based on whether Cloudflare blocks your network.
⚠️ Known Limitations
1. Cloudflare Bot-Management (Critical)
LinkedIn's Voyager API is protected by Cloudflare. On many networks and IP addresses, Cloudflare returns HTTP 302 redirect loops with Set-Cookie: __cf_bm that cannot be resolved with plain HTTP — even with:
- A valid
li_atcookie - A proper
JSESSIONID(CSRF token) - A full cookie jar capturing and replaying
__cf_bm
Result: All data tools return "Too many redirects (5)" instead of data.
Why: Cloudflare's bot detection evaluates TLS fingerprint, JavaScript execution, and browser behavior. Plain Node fetch() (and curl) cannot satisfy these checks.
Workarounds:
- Try from a different network/IP (residential IPs sometimes pass)
- Use the browser-based alternative (stickerdaniel/linkedin-mcp-server) which uses a real browser
- Deploy behind a residential proxy
2. OAuth Backs 0 Tools
All 36 tools use LinkedIn's internal Voyager API (voyagerGet/voyagerPost). The restGet/restPost methods (for OAuth's official REST API) have zero call sites. OAuth configuration is accepted but produces no working tools.
3. Raw Response Format
Voyager API responses are returned as-is: { data, included[] } with URN references unresolved. Responses can be large and are not optimized for LLM consumption.
⚙️ Environment Variables
| Variable | Required | Default | Description |
|---|---|---|---|
LINKEDIN_ACCESS_TOKEN | One of these | — | LinkedIn OAuth access token |
LINKEDIN_COOKIE | One of these | — | LinkedIn li_at session cookie |
LINKEDIN_CSRF_TOKEN | Yes (with cookie) | — | LinkedIn JSESSIONID cookie — required for CSRF protection |
PORT | No | 3000 | HTTP server port |
TRANSPORT | No | stdio | Transport mode: stdio or http |
LOG_LEVEL | No | info | Log level: debug, info, warn, error |
CACHE_TTL | No | 300 | Cache TTL in seconds |
RATE_LIMIT_RPM | No | 30 | Max requests per minute |
REQUEST_TIMEOUT | No | 30000 | Request timeout in milliseconds |
CORS_ORIGIN | No | localhost | Allowed CORS origin for HTTP transport |
🛠 Development
# Clone the repo
git clone https://github.com/devag7/linkedin-mcp.git
cd linkedin-mcp
# Install dependencies
npm install
# Run in development mode (stdio)
npm run dev
# Run in HTTP mode
npm run dev -- --transport http --port 3000
# Run tests
npm test
# Type checking
npm run typecheck
# Lint
npm run lint
# Build for production
npm run build
🤝 Contributing
Contributions are welcome! Please see our Contributing Guide for details.
- Fork the repository
- Create your feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'feat: add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
📄 License
This project is licensed under the MIT License — see the LICENSE file for details.
⚠️ Disclaimer
This project is not officially affiliated with LinkedIn. Use responsibly and in accordance with LinkedIn's Terms of Service. The authors are not responsible for any misuse or account restrictions.
📚 Related Resources
- 📘 Full Setup Guide — Step-by-step instructions with example prompts for all 36 tools
- 📥 Latest Release — Download Claude Desktop config file
- 🔧 MCP Protocol — Learn about the Model Context Protocol
- 📦 npm Package —
npm install linkedin-mcp-tools
⭐ Star this repo to support the project!
If LinkedIn MCP saves you time, show your support with a star. It helps others discover the most comprehensive LinkedIn MCP server available.
Made with ❤️ by Dev Agarwalla