LinkedIn MCP Server (Official API)

⚡ NOW USING OFFICIAL LINKEDIN API ⚡

Through this LinkedIn MCP server, AI assistants like Claude can connect to LinkedIn using the official LinkedIn API with proper OAuth 2.0 authentication. This ensures compliance with LinkedIn's Terms of Service and provides enterprise-grade reliability.

🔄 Migration Completed: Web Scraping → Official API

This server has been fully migrated from web scraping to the official LinkedIn API:

✅ What Changed:

• Web scraping → Official LinkedIn REST API

• Cookie authentication → OAuth 2.0 with access tokens

• Chrome/Selenium → HTTP API client

• Terms violation → Fully compliant API usage

⚠️ Important Limitations:

• Profile access: Only your own authenticated profile (not public profiles)

• Company access: Only companies you manage or have admin access to

• Job access: Only job postings from your companies

• No public search: Cannot search arbitrary profiles/companies/jobs

🎯 New Capabilities:

• OAuth 2.0 authentication flow

• Compliant LinkedIn API usage

• Enterprise-ready architecture

• Stable API contract (no breaking changes from UI updates)

Related MCP server: LinkedIn MCP Server

🚀 Quick Start with Official API

1. Prerequisites

• LinkedIn Developer Account

• LinkedIn Developer Application with OAuth credentials

2. Setup LinkedIn Developer App

# See detailed setup guide
cat LINKEDIN_PERMISSIONS_SETUP.md

3. Configure OAuth Credentials

export LINKEDIN_CLIENT_ID="your_client_id"
export LINKEDIN_CLIENT_SECRET="your_client_secret"
export LINKEDIN_REDIRECT_URI="http://localhost:8000/auth/callback"

# Optional: Direct access token for testing
export LINKEDIN_ACCESS_TOKEN="your_access_token"

4. Install and Run

# Install dependencies
pip install linkedin-api-client fastmcp keyring python-dotenv

# Run the server
python -m linkedin_mcp_server

5. OAuth Authentication Flow

1. Use get_oauth_authorization_url tool to get authorization URL

2. Direct user to authorization URL for consent

3. Get authorization code from callback

4. Use exchange_oauth_code to get access token

5. Use API with authenticated access token

Installation Methods

https://github.com/user-attachments/assets/eb84419a-6eaf-47bd-ac52-37bc59c83680

Usage Examples (Official API)

Help me set up OAuth authentication with LinkedIn

Get my LinkedIn profile information

What companies do I have admin access to on LinkedIn?

Show me job postings from companies I manage

What are the limitations of the official LinkedIn API?

Features & Tool Status (Official LinkedIn API)

☁️ Azure Functions Deployment (Recommended - Cloud)

Prerequisites: Azure CLI and Azure Functions Core Tools installed.

Quick Deployment

Run the automated deployment script:

# Make the script executable
chmod +x azure-deploy.sh

# Run deployment (will prompt for configuration)
./azure-deploy.sh

The script will:

1. Create Azure resource group

2. Set up Function App with consumption plan

3. Deploy the LinkedIn MCP Server

4. Configure environment variables

5. Provide access URLs

Manual Deployment

1. Login to Azure:

az login

2. Create resources:

# Create resource group
az group create --name linkedin-mcp-rg --location "East US"

# Create storage account  
az storage account create --name linkedinmcpstorage --location "East US" --resource-group linkedin-mcp-rg --sku Standard_LRS

# Create function app
az functionapp create --resource-group linkedin-mcp-rg --consumption-plan-location "East US" --runtime python --runtime-version 3.11 --functions-version 4 --name linkedin-mcp-server --storage-account linkedinmcpstorage

3. Deploy and configure:

# Deploy the function
func azure functionapp publish linkedin-mcp-server

# Set LinkedIn cookie
az functionapp config appsettings set --name linkedin-mcp-server --resource-group linkedin-mcp-rg --settings LINKEDIN_COOKIE="your_linkedin_cookie_here"

Getting the LinkedIn Cookie

1. Open LinkedIn and login

2. Open Chrome DevTools (F12 or right-click → Inspect)

3. Go to Application > Storage > Cookies > https://www.linkedin.com

4. Find the cookie named li_at

5. Copy the Value field (this is your LinkedIn session cookie)

6. Use this value as your LINKEDIN_COOKIE in the Azure Function settings

Usage

Once deployed, your LinkedIn MCP Server will be available at:

• Health Check: https://your-function-app.azurewebsites.net/api/health

• MCP Endpoint: https://your-function-app.azurewebsites.net/api/mcp

Client Configuration for Azure Functions:

{
  "mcpServers": {
    "linkedin": {
      "command": "curl",
      "args": [
        "-X", "POST", 
        "https://your-function-app.azurewebsites.net/api/mcp",
        "-H", "Content-Type: application/json",
        "-d", "@-"
      ]
    }
  }
}

Azure Functions Benefits

• Serverless: No server management required

• Auto-scaling: Automatically scales based on demand

• Pay-per-use: Only pay for actual executions

• High availability: Built-in redundancy and failover

• Integrated monitoring: Azure Application Insights included

• Secure: Function-level authentication and Azure security

Cost Estimation

Azure Functions consumption plan pricing:

• Free tier: 1 million requests and 400,000 GB-seconds per month

• Beyond free tier: $0.20 per million executions + $0.000016/GB-second

• Typical cost: $5-20/month for moderate usage (100-500 LinkedIn operations/day)

Environment Variables:

• LINKEDIN_COOKIE - LinkedIn session cookie (required)

• LINKEDIN_MCP_LOG_LEVEL - Logging level (DEBUG, INFO, WARNING, ERROR)

• LINKEDIN_MCP_HEADLESS - Run browser in headless mode (default: 1)

• LINKEDIN_MCP_LAZY_INIT - Initialize browser on first request (default: 1)

Monitoring:

# View live logs
az functionapp logs tail --name linkedin-mcp-server --resource-group linkedin-mcp-rg

# Enable Application Insights for detailed monitoring
az monitor app-insights component create --app linkedin-mcp-server --location "East US" --resource-group linkedin-mcp-rg

Common Issues:

• Function timeout: Default timeout is 5 minutes. Browser initialization takes 10-30 seconds on cold start.

• Memory limits: Chrome requires significant memory. Consider Premium plan for heavy usage.

• LinkedIn blocking: Use valid cookies and avoid excessive requests.

Debug locally:

# Install Azure Functions Core Tools
npm install -g azure-functions-core-tools@4 --unsafe-perm true

# Run locally
func start

# Test locally
curl http://localhost:7071/api/health

📦 Claude Desktop (DXT Extension)

Prerequisites: Claude Desktop installed

One-click installation for Claude Desktop users:

1. Download the DXT extension

2. Double-click to install into Claude Desktop

3. Set your LinkedIn cookie in the extension settings

Getting the LinkedIn Cookie

1. Open LinkedIn and login

2. Open Chrome DevTools (F12 or right-click → Inspect)

3. Go to Application > Storage > Cookies > https://www.linkedin.com

4. Find the cookie named li_at

5. Copy the Value field (this is your LinkedIn session cookie)

6. Use this value as your LINKEDIN_COOKIE in the configuration

DXT Extension Setup Help

Installation issues:

• Ensure you have Claude Desktop installed

• Download the latest DXT extension from releases

Login issues:

• Ensure your LinkedIn cookie is set and correct

• Make sure you have only one active LinkedIn session per cookie at a time. Trying to open multiple sessions with the same cookie will result in a cookie invalid error.

• LinkedIn may require a login confirmation in the LinkedIn mobile app

• You might get a captcha challenge if you logged in a lot of times in a short period of time, then try again later or follow the local setup instructions to run the server manually in --no-headless mode where you can debug the login process (solve captcha manually)

🚀 uvx Setup (Quick Install - Universal)

Prerequisites: Make sure you have uv installed.

Installation

Run directly from GitHub without cloning:

# Run directly from GitHub (latest version)
uvx --from git+https://github.com/stickerdaniel/linkedin-mcp-server linkedin-mcp-server --help

# Run with your LinkedIn cookie
uvx --from git+https://github.com/stickerdaniel/linkedin-mcp-server linkedin-mcp-server --cookie "li_at=YOUR_COOKIE_VALUE"

Getting the LinkedIn Cookie

1. Open LinkedIn and login

2. Open Chrome DevTools (F12 or right-click → Inspect)

3. Go to Application > Storage > Cookies > https://www.linkedin.com

4. Find the cookie named li_at

5. Copy the Value field (this is your LinkedIn session cookie)

6. Use this value as your LINKEDIN_COOKIE in the configuration

Run the server with the --get-cookie flag:

uvx --from git+https://github.com/stickerdaniel/linkedin-mcp-server \
  linkedin-mcp-server --get-cookie

Copy the cookie from the output and set it as LINKEDIN_COOKIE in your client configuration. If this fails with a captcha challenge, use the Chrome DevTools method above.

uvx Setup Help

Client Configuration:

{
  "mcpServers": {
    "linkedin": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/stickerdaniel/linkedin-mcp-server",
        "linkedin-mcp-server"
      ],
      "env": {
        "LINKEDIN_COOKIE": "li_at=YOUR_COOKIE_VALUE"
      }
    }
  }
}

Transport Modes:

• Default (stdio): Standard communication for local MCP servers

• Streamable HTTP: For web-based MCP server

CLI Options:

• --log-level {DEBUG,INFO,WARNING,ERROR} - Set logging level (default: WARNING)

• --no-lazy-init - Login to LinkedIn immediately instead of waiting for the first tool call

• --transport {stdio,streamable-http} - Set transport mode

• --host HOST - HTTP server host (default: 127.0.0.1)

• --port PORT - HTTP server port (default: 8000)

• --path PATH - HTTP server path (default: /mcp)

• --get-cookie - Attempt to login with email and password and extract the LinkedIn cookie

• --cookie {cookie} - Pass a specific LinkedIn cookie for login

• --user-agent {user_agent} - Specify custom user agent string to prevent anti-scraping detection

Basic Usage Examples:

# Run with cookie from environment variable
LINKEDIN_COOKIE="YOUR_COOKIE_VALUE" uvx --from git+https://github.com/stickerdaniel/linkedin-mcp-server linkedin-mcp-server

# Run with cookie via flag
uvx --from git+https://github.com/stickerdaniel/linkedin-mcp-server linkedin-mcp-server --cookie "YOUR_COOKIE_VALUE"

# Run with debug logging
uvx --from git+https://github.com/stickerdaniel/linkedin-mcp-server linkedin-mcp-server --log-level DEBUG

# Extract cookie with credentials
uvx --from git+https://github.com/stickerdaniel/linkedin-mcp-server linkedin-mcp-server --get-cookie

HTTP Mode Example (for web-based MCP clients):

uvx --from git+https://github.com/stickerdaniel/linkedin-mcp-server linkedin-mcp-server \
  --transport streamable-http --host 127.0.0.1 --port 8080 --path /mcp

Test with mcp inspector:

1. Install and run mcp inspector bunx @modelcontextprotocol/inspector

2. Click pre-filled token url to open the inspector in your browser

3. Select Streamable HTTP as Transport Type

4. Set URL to http://localhost:8080/mcp

5. Connect

6. Test tools

Installation issues:

• Ensure you have uv installed: curl -LsSf https://astral.sh/uv/install.sh | sh

• Check uv version: uv --version (should be 0.4.0 or higher)

Cookie issues:

• Ensure your LinkedIn cookie is set and correct

• Cookie can be passed via --cookie flag or LINKEDIN_COOKIE environment variable

• Make sure you have only one active LinkedIn session per cookie at a time

Login issues:

• LinkedIn may require a login confirmation in the LinkedIn mobile app for --get-cookie

• You might get a captcha challenge if you logged in a lot of times in a short period

🐍 Local Setup (Develop & Contribute)

Prerequisites: Chrome browser and Git installed

ChromeDriver Setup:

1. Check Chrome version: Chrome → menu (⋮) → Help → About Google Chrome

2. Download matching ChromeDriver: Chrome for Testing

3. Make it accessible:

   • Place ChromeDriver in PATH (/usr/local/bin on macOS/Linux)

   • Or set: export CHROMEDRIVER_PATH=/path/to/chromedriver

   • if no CHROMEDRIVER_PATH is set, the server will try to find it automatically by checking common locations

Installation

# 1. Clone repository
git clone https://github.com/stickerdaniel/linkedin-mcp-server
cd linkedin-mcp-server

# 2. Install UV package manager
curl -LsSf https://astral.sh/uv/install.sh | sh
uv python # install python if you don't have it

# 3. Install dependencies and dev dependencies
uv sync
uv sync --group dev

# 4. Install pre-commit hooks
uv run pre-commit install

# 5. Start the server once manually
# You will be prompted to enter your LinkedIn credentials, and they will be securely stored in your OS keychain
# Once logged in, your cookie will be stored in your OS keychain and used for subsequent runs until it expires
uv run -m linkedin_mcp_server --no-headless --no-lazy-init

Local Setup Help

CLI Options:

• --no-headless - Show browser window (debugging)

• --log-level {DEBUG,INFO,WARNING,ERROR} - Set logging level (default: WARNING)

• --no-lazy-init - Login to LinkedIn immediately instead of waiting for the first tool call

• --get-cookie - Login with email and password and extract the LinkedIn cookie

• --clear-keychain - Clear all stored LinkedIn credentials and cookies from system keychain

• --cookie {cookie} - Pass a specific LinkedIn cookie for login

• --user-agent {user_agent} - Specify custom user agent string to prevent anti-scraping detection

• --transport {stdio,streamable-http} - Set transport mode

• --host HOST - HTTP server host (default: 127.0.0.1)

• --port PORT - HTTP server port (default: 8000)

• --path PATH - HTTP server path (default: /mcp)

• --help - Show help

HTTP Mode Example (for web-based MCP clients):

uv run -m linkedin_mcp_server --transport streamable-http --host 127.0.0.1 --port 8000 --path /mcp

Claude Desktop:

{
  "mcpServers": {
    "linkedin": {
      "command": "uv",
      "args": ["--directory", "/path/to/linkedin-mcp-server", "run", "-m", "linkedin_mcp_server"]
    }
  }
}

Login/Scraping issues:

• Use --no-headless to see browser actions (captcha challenge, LinkedIn mobile app 2fa, ...)

• Add --no-lazy-init to attempt to login to LinkedIn immediately instead of waiting for the first tool call

• Add --log-level DEBUG to see more detailed logging

• Make sure you have only one active LinkedIn session per cookie at a time. Trying to open multiple sessions with the same cookie will result in a cookie invalid error. E.g. if you have a logged in browser session with a docker container, you can't use the same cookie to login with the local setup while the docker container is running / session is not closed.

ChromeDriver issues:

• Ensure Chrome and ChromeDriver versions match

• Check ChromeDriver is in PATH or set CHROMEDRIVER_PATH in your env

Python issues:

• Check Python version: uv python --version (should be 3.12+)

• Reinstall dependencies: uv sync --reinstall

Feel free to open an issue or PR!

Acknowledgements

Built with LinkedIn Scraper by @joeyism and FastMCP.

⚠️ Use in accordance with LinkedIn's Terms of Service. Web scraping may violate LinkedIn's terms. This tool is for personal use only.

Star History

License

This project is licensed under the Apache 2.0 license.