LinkedIn MCP Servers Compared, Honestly (Updated Monthly)

Last updated: 27 August 2026. This page is refreshed monthly. Tools in this space change quickly — features ship, waitlists open, pages get rewritten — so every claim below is dated, and anything we could not verify is marked as not verified rather than guessed at.

Kalovio is our own tool, so treat this page for what it is: a comparison written by one of the entrants. The way we keep it useful is by quoting competitors in their own words, linking to the page the quote came from, and refusing to describe a feature we have not confirmed. If you find something here that is out of date or unfair, it is a bug and we want to know.

Which LinkedIn MCP servers exist, and what does each actually do?

First, the term. MCP is the Model Context Protocol — an open standard that lets an AI assistant like Claude or ChatGPT talk to an outside tool. A “LinkedIn MCP server” is any tool that speaks that standard and touches LinkedIn somehow. That is a very wide net, and it is why searching for one returns things that have almost nothing in common.

Key to the table. “Not verified” means we could not confirm it from the tool’s own public page and will not guess. “OAuth” is the sign-in standard where you approve on LinkedIn’s own screen and can revoke access later from your settings. “Cookie session” means the tool borrows the logged-in session your browser already has. “Self-install” means you download and run it on your own machine; “hosted” means someone else runs the server.

Scrapers and data platforms: the ones that read, not write

This is where most search traffic lands, and it is the biggest source of confusion. These tools are good at what they do. What they do is not posting.

stickerdaniel/linkedin-mcp-server

This open-source project is the number one Google result for “linkedin mcp”, which means an enormous number of people meet it first and assume it is the category. It is not a posting tool. Its README says plainly that it “does not publish or create posts on your behalf”, and it also warns that accounts “can be restricted or banned”.

Credit where it is due: that is unusually honest documentation, and it tells you exactly what you are getting. It works by driving a browser with your logged-in session, which is how it reads profiles and search results that no self-serve API exposes. That approach is what produces both the capability and the warning. If you want to pull profile data into an assistant and you accept the risk the author has spelled out, this is the well-known option. If you wanted to write posts, it is the wrong tool and it says so itself.

Apify and Bright Data

Both are large data-extraction platforms that now offer LinkedIn MCP endpoints. You connect with a platform API key — a secret string that identifies your account with them, not with LinkedIn — and you get structured data back into your assistant.

These are built for research and lead-list work: scraping profiles, companies, job listings. They are not publishing tools, and they do not present themselves as such. If your job is “find me 200 companies matching this profile”, one of these is the right shape of answer. If your job is “write and schedule my posts”, neither will help.

Automation platforms: Zapier

Zapier’s LinkedIn MCP exposes only about two actions. That is a deliberate design choice, not a flaw — Zapier’s value has always been the workflow builder around a connection, not the depth of any single one. If you already run Zapier, and you want an assistant to trigger something simple as part of a larger chain, it is a sensible fit.

What you will not get is a content workflow: no drafts library, no queue, no voice profile, no media pipeline. You are getting a couple of actions inside a much bigger automation product. Judge it as that.

Posting tools: the ones that actually write

Taplio

Taplio is an established LinkedIn content product, and its MCP is an addition to that. Its own page states the MCP is “text-only... you cannot upload images, video, or PDF carousels”.

That is a real limit worth knowing before you subscribe, and it is to Taplio’s credit that they state it rather than leaving you to discover it. Whether it matters depends on you: if your LinkedIn posting is mostly text, the limit is irrelevant and you are buying into a mature product with a lot else in it. If image posts or PDF carousels are your format, the MCP will not do them today.

ContentIn

Worth mentioning only so you do not waste time looking for it: ContentIn’s LinkedIn MCP server does not exist yet. Their page says it is “currently in build, with a waitlist”. That is a fair way to gauge demand, but it is not something you can connect this week. We will move it into the main table when it ships.

Postiv and Carly

Both are posting tools in this space, and both are paid or trial-only. We have deliberately left most of their row blank rather than fill it from marketing copy. Neither page we checked names an OAuth scope or an API endpoint, which is also true of Taplio and ContentIn — none of the four ranking SaaS pages names a single scope or endpoint. That is not an accusation of wrongdoing; it is just a gap that makes an honest comparison hard. If you are evaluating them, ask them directly how they sign in to LinkedIn, and see the checklist at the end of this page.

Kalovio

Ours. Free public beta, no card, hosted at https://kalovio.com/mcp, and it works on Claude’s free plan. Sign-in is OAuth on LinkedIn’s own page, using two free self-serve products: Sign In with LinkedIn for identity, and Share on LinkedIn for the posting permission w_member_social. Your LinkedIn token is encrypted and never leaves the backend, and never reaches the assistant you are chatting with. Nothing is scraped.

The part we think matters most is the confirm gate. Publishing, scheduling, queueing and deleting are two-step: the first call returns a preview plus a signed token covering the exact wording, and nothing goes out until a second call arrives with your confirmation and a matching token. That is the practical defence against prompt injection — hidden instructions in content the model reads.

Now the limits, because they are the point of this page:

• It cannot read posts it did not create. LinkedIn’s r_member_social permission is closed to self-serve apps, so learning your voice only works from posts published through Kalovio.
• Multi-image and PDF posts ride LinkedIn’s versioned gateway, which may be product-gated for a self-serve app. A refusal is handled with a clear message, but we will not promise it works on every account — try it on yours.
• Tokens expire in about 60 days and there is no refresh token at this tier, so you reconnect periodically.
• No engagement analytics. That needs LinkedIn’s partner-gated Community Management API, which requires a registered legal company and app review. We do not have it. Link-click counts on our own short links are not the same thing.
• ChatGPT needs a paid plan, because Developer Mode is required, and it is web only. Claude works on the free plan.
• Free beta today. Pricing may change later, and we would rather say that than imply otherwise.

What about an official LinkedIn MCP server?

There is not one. LinkedIn has no official MCP server, so every tool on this page is a third party building on LinkedIn’s public APIs or, in the scrapers’ case, on a browser session. If a page implies it is the official one, that alone is a reason to look closer.

How should I pick one?

Start with the job, not the tool:

• You want to read LinkedIn data — profiles, companies, jobs. Look at Apify, Bright Data, or the stickerdaniel project, and read their own risk warnings before you connect an account you care about.
• You want to publish and schedule your own posts — you are choosing between Taplio, Postiv, Carly and Kalovio. Ask each one the same four questions below.
• You want one small action inside a bigger automation — Zapier.

The four questions worth asking any posting tool, whoever built it:

• Does connecting send me to linkedin.com? If setup asks for a cookie or a password, it is not OAuth, and you cannot revoke it from your LinkedIn settings afterwards.
• Which scope does it use? The answer for member posting is w_member_social. A tool that cannot name it is asking you to take its word.
• Do I approve the exact text before it publishes? With an AI assistant in the loop, this is the difference between a tool and an accident.
• What can it not do? Any answer that includes no limits at all is a sales page, not an answer.

When this page changes

Monthly, and sooner if something material moves — a waitlist opens, a limit is lifted, a page we quoted is rewritten. Quotes are dated so you can check them yourself against the linked source. If a tool believes its row is wrong, the fastest fix is to point us at the public page that says otherwise, and we will update the row and the date.