We open-sourced our LinkedIn automation β it's an MCP server (free, MIT)
We just shipped something a lot of you will get use out of: an open-source LinkedIn MCP server. It plugs straight into your AI client (Claude Code, Cursor, Claude Desktop) and lets it act on LinkedIn through your own logged-in Chrome β no third-party API, no per-seat fees, nothing sitting between you and LinkedIn. What it can do: β Reply to your DMs β whole inbox, drafted + sent from chat β Send connection requests β personalised, batched, daily caps enforced β Scrape every commenter on any post β turn a viral thread into a lead list β Post to your feed β from a single prompt β Comment on posts at scale β top-level or in reply threads β Look up any profile β headline, work history, network distance Why it's different from other LinkedIn tools: β No third-party API. It runs inside your real Chrome with your real cookies β indistinguishable from clicks you'd make yourself. β MIT licensed. Fork it, ship it, embed it in your own stack. β No pasting li_at cookies into janky tools. You log into Chrome once. β Built-in safety: send verification (throws if a DM didn't actually land), hard daily caps, single-writer lockfile. Repo (public, ~10-min setup β README has everything): https://github.com/Michaelrecycle/linkedin-mcp How you actually use it β just talk to Claude: β’ "Triage my LinkedIn inbox and draft replies to anyone asking about pricing." β’ "Scrape everyone who commented on my last post and pull their headlines into a list." β’ "Send a connection request to these 10 people, personalised note each." β’ "Post this to my feed, then check the comments in an hour." It's unofficial and a ToS gray zone (the README is honest about that), so keep it human-paced. But if you're running outreach, this replaces a $50β200/mo tool with something you fully own. Questions? Drop them below π