ModuleNotFoundError: 'linkedin_scraper.exceptions' during installation

[Manamama]

ModuleNotFoundError: 'linkedin_scraper.exceptions' during installation

Problem Description

When attempting to run linkedin-mcp-server, a ModuleNotFoundError occurs for linkedin_scraper.exceptions, even after linkedin-scraper appears to be successfully installed via pip. This prevents linkedin-mcp-server from functioning.

Steps to Reproduce

1. Ensure linkedin-mcp-server is installed (e.g., pip install /path/to/linkedin-mcp-server).
2. Attempt to run the main.py script from linkedin-mcp-server:

   python3 /path/to/linkedin-mcp-server/main.py

Expected Behavior

linkedin-mcp-server should start and run without module import errors, assuming all dependencies are correctly installed.

Actual Behavior

The following error is raised:

ModuleNotFoundError: No module named 'linkedin_scraper.exceptions'

Further investigation using pip show -f linkedin-scraper reveals that exceptions.py is missing from the installed linkedin_scraper package, despite being present in the source repository.

Workaround/Resolution

The issue can be resolved by manually cloning the linkedin_scraper repository and installing it from the local path:

1. Uninstall any existing linkedin-mcp-server and linkedin-scraper installations:

   pip uninstall -y linkedin-mcp-server
   pip uninstall -y linkedin-scraper

2. Clone the linkedin_scraper repository:

   git clone https://github.com/stickerdaniel/linkedin_scraper.git

3. Install linkedin-scraper from the local clone:

   pip install ./linkedin_scraper

4. Reinstall linkedin-mcp-server from its local path:

   pip install /path/to/linkedin-mcp-server

Thanks to this workaround, the linkedin-mcp-server is now successfully running!

Environment

• Operating System: Linux
• Python Version: 3.10 (as inferred from cpython-310.pyc in pip show output)
• pip Version: (latest as of July 12, 2025)
• linkedin-mcp-server Version: 1.2.6
• linkedin-scraper Version: 2.11.5 (as per pyproject.toml and pip show)

Additional Notes

It appears that the uv package manager, as suggested in the README.md for linkedin-mcp-server, is completely unnecessary for a successful installation and may even contribute to confusion regarding dependency resolution. It is recommended to consider removing uv from the installation instructions.

(To the tune of "Twinkle, Twinkle Little Star")

Code was broken, quite a fright,
Modules missing, day and night.
But with a plan, and careful hand,
We fixed the errors, through the land!
LinkedIn server, now you gleam,
Working smoothly, like a dream!

-- Gemini CLI on behalf of https://github.com/Manamama

[Manamama]

Human's update, pro memoria:

Also this:
cp ~/linkedin-mcp-server/main.py /home/user/.local/lib/python3.10/site-packages/linkedin_mcp_server/main.py

and then in JSON settings:

    "linkedin": {
      "command": "python",
      "args": [
        "/home/user/.local/lib/python3.10/site-packages/linkedin_mcp_server/main.py"
      ],
      "env": {
        "LINKEDIN_COOKIE": "$LINKEDIN_COOKIE"
      }

then indeed:

/Downloads/GitHub$ gemini
MCP STDERR (linkedin): 

╭─ FastMCP 2.0 ────────────────────────────────────────────────────────────────╮
│                                                                              │
│                                                                              │
│         _ __ ___ ______           __  __  _____________       ____           │
│     ____                                                                     │
│        _ __ ___ / ____/___ ______/ /_/  |/  / ____/ __ \     |___ \  /       │
│     __ \                                                                     │
│       _ __ ___ / /_  / __ `/ ___/ __/ /|_/ / /   / /_/ /     ___/ / / /      │
│     / /                                                                      │
│      _ __ ___ / __/ / /_/ (__  ) /_/ /  / / /___/ ____/     /  __/_/ /_/     │
│     /                                                                        │
│     _ __ ___ /_/    \__,_/____/\__/_/  /_/\____/_/                           │
│     /_____(_)____/                                                           │
│                                                                              │
│                                                                              │
│                                                                              │
│     🖥️  Server name:     linkedin_scraper                                     │
│     📦 Transport:       STDIO                                                │
│                                                                              │
│     📚 Docs:            https://gofastmcp.com                                │
│     🚀 Deploy:          https://fastmcp.cloud                                │
│                                                                              │
│     🏎️  FastMCP version: 2.10.4                                               │
│     🤝 MCP version:     1.10.1                                               │
│                                                                              │
│                                                                              │
╰──────────────────────────────────────────────────────────────────────────────╯

One could patch it or script it.

[stickerdaniel]

Hey, try the exact installation instructions or use the containerized Docker version. (This MCP is also released in the Docker MCP Catalog)
This repo uses a custom fork of the linkedin_scraper package, which I've modified to address some issues. The PR to the original repo is still open. If you use uv sync, the correct linkedin_scraper package from the pyproject toml will be installed, and you shouldn't experience this issue.

[Manamama]

Uhm. Human here this time, so messy off-the-cuff thoughts instead.

Containerized docker started to download some 300 MBs, I quickly stopped it (I never used it before, maybe that is why). Now, uv, venv and mambas and such I detest as they eat up my Notebook's disk space.
To wit: yours now runs and says hello to LinkedIn, but indeed, it says that the cookie (found via "Cookie Editor") is wrong (session with LinkedIn refreshed twice), so you may be on to smth.

I think I will go the way of below then (yes, I know, completely different architecture of "reading" LinkedIn stuff, but then almost a one liner and R/W to boot):

    "browsermcp": {
      "command": "npx",
      "args": [
        "@browsermcp/mcp@latest"
      ]
    },

Ciao and thanks for the work!