The LinkedIn MCP server that can publish
Liftli is a remote LinkedIn MCP server. Paste one URL into Claude, ChatGPT, Cursor or Grok, sign in with LinkedIn, and your AI can plan, write, publish and schedule LinkedIn posts in your own voice. Everything goes through LinkedIn's official API, and nothing ships until you approve it.
What a LinkedIn MCP server does
A LinkedIn MCP server is a connector that lets an AI assistant work with LinkedIn through the Model Context Protocol (MCP). Instead of copying a draft out of a chat window and into LinkedIn, you ask the assistant to "post this tomorrow at 9" and it does, after you say yes.
Most LinkedIn MCP servers you will find on GitHub are built to read LinkedIn: look up profiles, pull job listings, fetch posts. Liftli is built to help you show up on LinkedIn: decide what to say, write it the way you write, and put it on your profile at the right time.
LinkedIn has no official MCP server of its own, so every option, Liftli included, comes from a third party. The comparison below shows how the main ones differ. New to using AI for LinkedIn? The Claude for LinkedIn guide starts with the simpler options: plain chat with a voice file, and free agent skills.
Connect it in a minute
One remote MCP URL works in every client. There is nothing to install, clone or run locally:
https://mcp.liftli.ai/mcp| Client | How to add the LinkedIn MCP |
|---|---|
| Claude (web, desktop, mobile) | Customize (briefcase icon) → Connectors → Add custom connector → paste https://mcp.liftli.ai/mcp |
| Claude Code | claude mcp add --scope user --transport http liftli https://mcp.liftli.ai/mcp |
| ChatGPT (custom app) | On a computer: Plugins → Add → Create MCP App (or + → Create app) → add the same URL. If there is no create option or ChatGPT asks for Developer mode, turn it on under Settings → Security and login |
| Codex, Cursor (beta) | Add a remote MCP server pointing at the same URL |
| Grok (grok.com) | grok.com/connectors → New Connector → Custom → the same URL |
Then say "connect my LinkedIn". Liftli sends you to LinkedIn's own sign-in page; you approve there, and LinkedIn hands Liftli permission to post for you. Liftli never sees your LinkedIn password. Full setup for every client is on the developers page.
What your AI can do on LinkedIn with Liftli
- Publish and schedule posts. "Post this now" or "schedule it for Tuesday at 8:30". Move, replace or cancel anything in the queue from the same chat.
- Write in your voice. Liftli extracts your writing voice from your own posts, so drafts sound like you and not like a chatbot.
- Pressure-test every draft. Each post runs through a plan, draft, critique and revise loop before it reaches you.
- Turn real work into posts. Voice notes, call transcripts, GitHub activity and chats become post ideas grounded in what actually happened.
- Draft sharp comments. Paste someone else's post and get one value-first comment that borrows their audience. You post it yourself.
- Rewrite your profile. Headline, About and Featured, rewritten as a landing page for the people you want to reach.
- Learn from results. Log how a post did and Liftli adjusts the plan. On Business, it sorts the people who engaged, from a list you paste, by fit with your ideal customer.
It remembers your strategy, voice and results between sessions and across clients, so Claude on your laptop and ChatGPT on your phone work from the same plan. See how it works for the full loop.
LinkedIn MCP servers compared
LinkedIn MCP servers reach LinkedIn in one of two ways. Some use LinkedIn's official API, with your permission given on LinkedIn's own sign-in page. Others act through your logged-in LinkedIn session, which lets them read and do much more, and which LinkedIn's rules on automation prohibit. Here is how the servers that come up most in search compare. Each name links to the project's own page, which is the source for its row.
| Server | Hosting | How it reaches LinkedIn | What it does | Posts for you? |
|---|---|---|---|---|
| Liftlithis page | Remote. Paste https://mcp.liftli.ai/mcp and sign in |
LinkedIn's official API, approved by you on LinkedIn's own sign-in page | Plans what to post, writes in your voice, critiques every draft, drafts comments, rewrites your profile | Yes. Publishes and schedules to your profile after your one-tap approval |
| stickerdaniel/ |
Self-hosted: uvx, Docker or a Claude Desktop bundle | Runs a real Chromium browser with your LinkedIn session cookies | Reads profiles, companies, jobs and your feed; reads and sends messages; sends and accepts connection requests | No post tool listed. Its own FAQ warns that LinkedIn's User Agreement prohibits automated access and that accounts can be restricted or banned |
| LinkMCPhosted service | Hosted. Sessions run on LinkMCP's servers | Your own LinkedIn account. Its site says it does not use an official LinkedIn partner API | People and Sales Navigator search, reading posts and who engaged, messages and InMail, connection requests, email and phone lookup (35 tools, per its site) | Yes: publishes, comments and reacts. Its FAQ says a restriction cannot be ruled out completely, so it paces actions with daily limits |
| Taplio MCPhosted, for Taplio accounts | Hosted at mcp.taplio.com. Needs a Taplio account (it offers a free trial) |
Your Taplio account's LinkedIn connection; its page says no scraping. Whether that is LinkedIn's official API is not stated on its MCP page (unverified) | Research, drafts, scheduling and analytics, plus a toolset that finds posts to comment on and drafts comments and replies (27 tools, per its page) | Yes, after you confirm |
| CData MCP Server for LinkedIncommercial, free trial | A server you run locally, or a hosted endpoint through CData Connect AI | LinkedIn's API through OAuth (its docs list the OAuth settings) | Exposes LinkedIn data as tables to query: company pages, follower and page statistics, company updates, comments. Built for reporting and analysis | Not its focus. Posting to a personal profile: unverified |
| felipfr/ |
Self-hosted Node.js project you build and run | The README describes a LinkedIn API client with token management, but not how you get the token (unverified) | Profile search and lookup, job search, messages to your connections, network statistics | No post tool listed. A small project: 4 commits when checked |
Checked against each project's own page on 14 October 2026. These tools change quickly, so open the link before you choose. Prices are left out on purpose: compare them on each project's own site. LinkedIn itself publishes no MCP server; its closest first-party option is a plugin inside ChatGPT that looks up professionals and does not post.
Which one should you pick
- You want to post on your personal profile from Claude or ChatGPT, in your own voice, without risking the account: Liftli. It uses the official API, waits for your approval, and plans and critiques the post instead of only sending it.
- You already run your LinkedIn through Taplio: the Taplio MCP brings your Taplio drafts, queue and comment workflow into your AI.
- You want to research profiles, jobs or your inbox from your own machine: stickerdaniel's server is the most popular open-source option (3.9k GitHub stars when checked). Its README is upfront about the account risk; take it seriously.
- You run outreach at volume (search, messages, contact lookup) and want it hosted: LinkMCP, with the restriction risk its own FAQ describes.
- You need company page numbers in a reporting workflow: CData.
- You want a small codebase to read or fork: felipfr/linkedin-mcpserver.
Not sure you need an MCP server yet? Claude can already write strong LinkedIn posts with the right inputs. The Claude for LinkedIn guide shows how, and the free LinkedIn agent skills teach your agent the craft before you connect anything.
Liftli compared with a self-hosted LinkedIn MCP
| Liftli | Typical open-source LinkedIn MCP | |
|---|---|---|
| Main job | Plan, write, publish and schedule your posts | Read profiles, jobs and posts |
| How it reaches LinkedIn | LinkedIn's official API, connected with LinkedIn's sign-in | Varies: check whether it uses the official API or your browser session |
| Setup | Paste one URL | Clone, install and run it on your machine |
| Posts to your profile | Yes, after your one-tap approval | Varies; many are read-only |
| Writes in your voice | Yes, from your own posts | No |
| Works on your phone | Yes, anywhere your AI client runs | Only where you run the server |
Self-hosted servers are a good fit if you want to research LinkedIn data from your own machine. If you want to publish consistently without putting your account at risk, a server that uses the official API is the safer route.
Using a LinkedIn agent skill? Add the MCP to post
Agent skills are instructions your AI runs locally. They can help it write a better LinkedIn post, but a skill cannot publish to your profile on its own. Liftli ships both: 31 free LinkedIn agent skills you can install with npx skills add liftli-ai/linkedin-agent-skills, and the MCP server that takes the finished post and publishes or schedules it.
Safe for your LinkedIn account
- Official API only. No browser extension, no bot actions on your profile.
- You approve every post. Nothing is published or scheduled without your explicit go-ahead.
- You stay in control. You can remove Liftli's access at any time from LinkedIn's settings.
Details on what Liftli can access and store are on the security page.
Pricing
Start free with no credit card: 3 posts (edit each one as much as you like), 10 comments, voice extraction, a starter strategy, and publishing and scheduling to LinkedIn from chat. Pro is $29 a month at the founding rate (it rises to $49 after the first 150 members, and your rate is kept) for unlimited LinkedIn posts; Business adds X and Substack. Liftli adds no AI markup, because it runs inside the AI subscription you already pay for. See all plans.