LinkedIn MCP Server: Free, Official API, Actually Posts (2026)

Search “LinkedIn MCP server” and you get a wall of GitHub repositories, a few paid products and one waitlist. Almost none of them tell you the two things you need to know first: whether LinkedIn made any of them (no), and whether the one you are looking at can actually write a post (usually no). This page is the complete answer.

What is an MCP server, in one paragraph?

MCP stands for Model Context Protocol — an open standard, published by Anthropic, for plugging AI assistants into outside tools. Think of it as a universal socket: instead of every AI app building its own custom LinkedIn integration, a service exposes one MCP server and any MCP-speaking assistant can use it. A remote MCP server is one that already runs on the internet, so you do not install anything — you paste a web address into your AI app, sign in once, and the assistant gains a new set of abilities. Claude and ChatGPT both support this today.

Is there an official LinkedIn MCP server?

No. As of August 2026, LinkedIn and its parent Microsoft do not ship, host or endorse an MCP server. There is no entry for one in LinkedIn’s developer product list. Every LinkedIn MCP server you can find is built by somebody else.

That is not automatically a problem. MCP is an open protocol, so LinkedIn’s participation was never required. The question that actually matters is not “is this official?” — none are — but what is it built on top of? There are only two real answers: LinkedIn’s public, documented API with your own consent, or scraping, which means logging in as you with a copied session cookie.

What is Kalovio?

Kalovio is a free, hosted, multi-tenant LinkedIn MCP server at https://kalovio.com/mcp. Multi-tenant just means many people share one running server while each account stays completely separate. You add it once to Claude or ChatGPT, sign in with LinkedIn on LinkedIn’s own consent screen, and then run your LinkedIn account by typing in plain English: write a post about X, attach these three images, schedule it for Tuesday morning.

It is built entirely on LinkedIn’s official self-serve API. It never scrapes, never asks for your password, and never asks you to paste a cookie.

Writing

• Draft posts in chat, then create, list, read, update and delete saved drafts.
• A per-user voice profile — you describe how you write once, and every later draft follows it.
• Posts run up to 3,000 characters, which is LinkedIn’s own hard limit.

Publishing and scheduling

• Publish immediately, or schedule for any date and time in your own timezone.
• A weekly posting queue: set your preferred slots once (for example Tuesday and Thursday at 9am), then drop posts into the next free slot without picking dates each time.
• List everything scheduled, cancel a scheduled post, list what Kalovio has published, and delete a post Kalovio published.

Media

• A single image.
• A multi-image carousel of 2 to 9 swipeable images.
• A document carousel — a PDF, PowerPoint or Word file, the format people usually call a “PDF carousel”.
• An image library with a 30-day retention, up to 100 files per user, 20 MB per file. You add images through an upload panel that appears in the chat, or through a plain browser upload page.
• A caption is optional when a file is attached, so an image-only post is a first-class post.

Tagging and tracking

• Clickable @mentions of company pages, resolved through a shared directory of company IDs Kalovio keeps.
• Clickable mentions of other Kalovio users, via a short collab code they share with you.
• Link-click tracking using kalovio.com/r/ short links, so you can see how many people clicked the link in a post.

Account

• Account status at a glance, and a one-click reconnect link when your LinkedIn access expires.

That is 26 tools in total, all reachable by describing what you want rather than clicking through a dashboard.

How do I know it really uses LinkedIn’s official API?

“Official API, no scraping” is a sentence anyone can type onto a landing page. None of the four LinkedIn MCP product pages currently ranking name a single permission scope or endpoint. Here is Kalovio’s, in full, so you can check it against LinkedIn’s own documentation.

• Identity: “Sign In with LinkedIn (OIDC)” — OIDC (OpenID Connect) is the standard way an app confirms who you are without ever seeing your password. This is what creates your Kalovio account.
• Posting permission: the “Share on LinkedIn” product, which grants the scope w_member_social. A scope is a single named permission you approve on LinkedIn’s consent screen. This one, and only this one, allows posting on your behalf.
• Endpoints: text posts, single-image posts and article shares go to POST /v2/ugcPosts. Multi-image carousels and document carousels use LinkedIn’s modern versioned API — /rest/images, /rest/documents and /rest/posts — sent with the headers LinkedIn-Version: 202606 and X-Restli-Protocol-Version: 2.0.0.
• Both products are free and self-serve. No partner status, no payment, no sales call. The gate is app approval, not money.

Two more things about how your access is held:

• Your LinkedIn token never leaves the backend. It is envelope-encrypted with AES-256-GCM (a strong standard cipher, with the key itself encrypted by a separate master key) and stored in the database. It is never handed to Claude or ChatGPT, never written to a log, never included in a tool’s output.
• Two separate sign-ins, never bridged. Your AI app signs in to Kalovio using OAuth 2.1. Kalovio separately signs in to LinkedIn using classic OAuth 2.0. The two are deliberately kept apart, because passing an AI app’s token upstream to LinkedIn would be both a spec violation and a security hole.

How does the LinkedIn MCP landscape compare?

The options fall into three groups, and they are built for genuinely different jobs. All quotes below are taken verbatim from each project’s own page.

The pattern is worth saying plainly: the most popular result for this search is a reader, not a writer. If your goal is to publish, most of the landscape is solving a different problem.

How do I connect it to Claude or ChatGPT?

Claude. Open Settings, then Connectors, then “Add custom connector”, and paste https://kalovio.com/mcp. Approve the connection, then sign in with LinkedIn on the consent screen that appears. This works on Claude’s free plan.

ChatGPT. You need Developer Mode, which is on the paid plans (Plus, Pro, Team, Enterprise, Edu) and on the web only. Go to Developer Mode, then Plugins, then the plus button, then New Plugin, set Auth to OAuth and paste the same address. One extra step catches people out: a Developer Mode connector must also be switched on inside each conversation from the tools “+” menu. It is not automatically active in a new chat.

Connecting inside ChatGPT does not carry over to Codex. Each app connects and signs in on its own.

What can it not do?

This section is longer than most competitors’ entire limitations section, on purpose. These are constraints of LinkedIn’s public API, not bugs, and no tool on the self-serve rails escapes them.

• It cannot read posts it did not create. The permission for that, r_member_social, is closed to self-serve apps. Kalovio can only see posts published through Kalovio. Anything claiming to read your whole history is scraping.
• Access expires in about 60 days. Self-serve apps get no refresh token, so you will reconnect periodically. Kalovio’s reconnect_linkedin hands you the link.
• Company tagging needs a numeric ID. LinkedIn gives self-serve apps no name-to-ID lookup, so Kalovio keeps a shared directory of companies it has already resolved. A company nobody has tagged yet may need its ID once.
• No real engagement analytics. Impressions, follower growth and reactions need LinkedIn’s Community Management API, which is partner-gated behind a registered legal company and app review. Kalovio offers link-click tracking, which is a different and much smaller thing. It is not LinkedIn engagement data.
• Multi-image and document posts ride LinkedIn’s versioned gateway, which may be product-gated for a self-serve app. If LinkedIn refuses, you get a clear message rather than a silent failure. Do not count on these two formats working on every account until you have tried them on yours.
• No DMs, no connection requests, no lead scraping. Kalovio does not do outreach of any kind, and that is deliberate.
• Chat image attachments do not reach the connector. Neither Claude nor ChatGPT forwards a file you dropped into the chat to a plain MCP server. You upload through Kalovio’s panel or upload link instead.
• It is a free public beta. No card required. Pricing may change later, and saying otherwise would be a promise nobody can keep.

Who is this for?

If you want to read profiles, companies or job listings in bulk, use a scraper or a data platform — and read their own warnings about account risk before you do. If you want to write, schedule and publish your own LinkedIn posts by talking to an AI assistant, you need something built on w_member_social, and there are not many of those. Kalovio is free while in beta, names exactly what it is built on, and shows you every post before it goes out.