LinkedIn MCP Server

LinkedIn MCP Server with Anthropic's Claude Integration

A Python-based MCP (Model Context Protocol) server that gets stuff from your LinkedIn profile and integrates with the Anthropic API for potential analysis tasks.

TL;DR Install for Claude Desktop/Code Access to the LinkedIn profile

Features

• Job Search URL Generation: Create properly formatted LinkedIn job search URLs with location, distance, and query parameters
• Job ID Retrieval: Extract job IDs from LinkedIn search pages with pagination support
• Job Metadata Extraction: Get detailed job information including title, company, description, and requirements
• CV Adaptation: Combined with the cv MCP Server, adapt Francisco's CV to match specific job requirements
• Built with FastMCP for high performance and with Pixi for dependency management and task running
• Source code organized in the src/ directory
• Includes configurations for:
  • Docker (optional, for containerization)
  • Linting (Ruff, Black, iSort)
  • Formatting
  • Type checking (MyPy)

Prerequisites

• Python 3.11+
• Pixi (for dependency management and task execution)
• uv (for building the MCP bundle)
• Docker (optional, for containerization)

Project Structure

This project follows the src layout for Python packaging.

Setup and Installation

1. Clone the repository (if applicable) or ensure you are in the project root directory.
2. Install dependencies using Pixi:

This command will create a virtual environment and install all necessary dependencies:

Running the Server

Pixi tasks are defined in pyproject.toml:

mcps (MCP Server)

Development Mode (with auto-reload)

The server will start at http://localhost:10000. It will automatically reload if you make changes to files in the src/ directory.

MCP Inspection Mode

This starts the inspector for the MCP Server.

Web scrapper

Development Tasks

Run Tests

Lint and Check Formatting

Apply Formatting and Fix Lint Issues

Build the Python Package

Creates sdist and wheel in dist/:

Docker Support (Optional)

Build the Docker Image

Run the Docker Container

TODO: Rewrite this if necessary. Docker support not yet done.

MCP Server Installation

Local Installation with MCP

Init the dxt

Init dxt project with a manifest

Note When creating the manifest, in the mcp_config section, put the full path to the python interpreter -> "command": "/Users/fperez/.pyenv/shims/python"

Bundle Python libs and Package Project as dxt

The output file linkedin-mcp-fps.dxt is created on the dxt-package directory. Alternatively, download the linkedin-mcp-fps.dxt file from releases (TODO).

With the packaged extension:

1. Double-click the .dxt file to install it in Claude Desktop
2. Alternatively, drag and drop it to Claude Desktop Settings/Extensions section
3. Restart Claude Desktop (In new Claude versions it's not necessary)
4. The extension should appear in your MCP servers list

Extension Requirements

This extension requires Python 3.11+ and includes all necessary dependencies bundled.

Dev Local Installation for Claude Desktop/Code (without DXT)

Remote Configuration for Claude Desktop/Code

For connecting to a remote MCP server:

Note: Update the host and port as needed for your deployment.

Currently I'm using render.com to host the MCP server. The configuration for Claude is in the config/claude.json file. It uses sse but it is deprecated now. TODO: Make streamable-https the default for remote.

Render requires requirements.txt to be present in the root directory. You can generate it using:

Also requires runtime.txt to be present in the root directory with the Python version specified:

Remember also to set the environment variables in the render.com dashboard:

User Guide

Available Tools

1. get_url_for_jobs_search: Generate LinkedIn job search URLs
2. get_new_job_ids: Get new job IDs from LinkedIn
3. get_jobs_raw_metadata: Extract detailed job information
4. adapt_cv_to_latest_job: Adapt CV to job requirements

After installing the MCP server, you can access its functionality in Claude Desktop/Code using the tools to get information about jobs in Linkedin. Combined with the functionality provided by the MCP Server serving my CV you can ask things like this:

or simply:

or, as a recruiter, get your posted LinkedIn job id and write:

Cache

The system uses a local cache to avoid re-scraping jobs:

• Cache location: ~/.linkedin-mcp/raw_job_description_cache/
• Format: JSONL (JSON Lines)
• Automatically managed
• In-memory cache for fast lookups
• Configurable cache keys and storage locations

Troubleshooting

If you encounter issues:

1. Import errors: Ensure all required Python packages are installed
2. WebDriver issues: Make sure Chrome is installed for Selenium
3. Connection errors: Check your internet connection for LinkedIn access
4. Permission errors: Ensure the cache directory is writable
5. Python path issues: Verify the manifest.json uses the correct Python executable path

Dev

Requirements

• pixi
• uv

Support

For issues and feature requests, visit: https://github.com/francisco-perez-sorrosal/linkedin-mcp

License

This project is licensed under the MIT License. See pyproject.toml (See LICENSE file) for details.