// the find
stickerdaniel/linkedin-mcp-server
Open-source MCP server for LinkedIn. Give Claude and any MCP-compatible AI agent access to profiles, companies, jobs, and messages.
An MCP server that gives AI assistants access to LinkedIn via browser automation — scraping profiles, companies, jobs, and messages through a real Chromium session controlled by Patchright. It's for developers building AI agents that need to interact with LinkedIn data without using an official (non-existent) API. The tool table is refreshingly honest: several features like send_message and connect_with_person are listed with open issue links rather than 'working'.
The honesty about what's broken is genuinely good — the status table links directly to issues rather than pretending everything works. Multiple deployment paths (uvx, Docker, local, .mcpb) are well-documented with specific troubleshooting steps that reflect real operational experience. Tool calls are serialized through a queue to protect the shared browser session, which is the right call for a single-user Playwright/Patchright setup. The .agents/ directory with SKILL.md files suggests the maintainer is already thinking about automated triage and reproduction — not just shipping and abandoning.
This is scraping LinkedIn's DOM, which means any UI change on LinkedIn's side breaks tools with zero warning — the maintainer can't version-control LinkedIn's frontend. LinkedIn's ToS explicitly prohibits automated access, so there's a real account-ban risk that's disclosed but can't be engineered away. The session model is inherently fragile: cookies expire, captchas appear, 2FA kicks in, and Docker users have to re-login on the host and remount — this gets tedious fast in production-like workflows. There's no rate limiting or request pacing built into the tool layer, so an aggressive agent can trigger LinkedIn's bot detection immediately.