LinkedIn Plugins for AI Assistants: What Exists Today

If you are shopping for a way to plug LinkedIn into an AI assistant, the listings all look similar. Every one of them says “LinkedIn integration”. Almost none of them say what the integration actually does, and the gap between two tools with identical descriptions can be the difference between reading someone’s profile and publishing a post from your own account.

So this page groups what exists by capability, not brand. Four groups, one question that splits them, and a check you can run on any tool in about a minute.

What does “LinkedIn plugin” even mean now?

The word is a leftover. “Plugin” came from the original ChatGPT plugin store, which no longer works the way it did, but the word stuck because it is what people type when they mean “an add-on that lets my AI assistant do something with LinkedIn”.

Today the mechanism behind almost all of these is MCP — the Model Context Protocol, an open standard for connecting AI assistants to outside tools. In Claude they are called custom connectors; in ChatGPT they are added through Developer Mode. Different words, same plumbing. What matters is not the label on the box but which of the four groups below the thing inside belongs to.

One fact worth stating up front, because it removes a whole category of confusion: LinkedIn publishes no official plugin, connector or MCP server of its own. Everything on the market is built by someone else, either on top of LinkedIn’s public API or by scraping the website. That distinction is the rest of this article.

What kinds of LinkedIn plugin exist right now?

Group 1: read-only scrapers

These are the most numerous and the easiest to find. They sign in as you — usually by asking you to paste a session cookie, the small string your browser holds that proves you are already logged in — and then read pages the way a browser does. They fetch profiles, company pages, search results, and sometimes your feed.

The best-known example is also the top Google result for “linkedin mcp”: stickerdaniel/linkedin-mcp-server. It is refreshingly honest about both of the things that define this group. Its README states plainly that it “does not publish or create posts on your behalf”, and it warns that accounts “can be restricted or banned”.

Read those two lines together, because they are the whole group in miniature. A scraper cannot post, and using one puts your account in a position LinkedIn does not sanction. That is not a criticism of the project — it documents its own limits better than most paid products do. It is simply what this category is.

Group 2: data-extraction platforms

Bigger, commercial versions of the same idea. Apify and Bright Data both offer LinkedIn MCP endpoints, and both are aimed at data extraction, not posting. You point them at profiles, companies or job listings and they return structured data at volume.

These are genuinely useful if your job is research, list building or market analysis, and they are honest about being extraction services. But if what you wanted was “let my AI assistant write and publish my LinkedIn posts”, this group answers a different question entirely. Buying one and expecting to publish is the most common mistake in this market.

Group 3: thin automation-platform integrations

The large automation platforms — the ones that wire hundreds of apps together — have started exposing their app catalogues over MCP. Zapier is the clearest case: its LinkedIn MCP exposes only about two actions.

Two actions is not a flaw in the design. Automation platforms are built for one-step triggers, so the LinkedIn surface they expose is deliberately narrow. It works if your need is narrow too. It falls apart the moment you want drafts, a schedule, a queue, images, a document post, or an approval step, because none of those are actions in the catalogue.

Group 4: posting tools built on the official API

The smallest group. These use LinkedIn’s own OAuth sign-in and the w_member_social permission — the named scope, meaning the specific permission you grant, that allows an app to publish on your behalf and nothing else — then post through POST /v2/ugcPosts or LinkedIn’s newer versioned /rest/posts endpoint.

Even inside this group the capabilities vary a lot, and vendors are not always forthcoming about where the edges are. Taplio’s own page is unusually direct about its: its MCP is “text-only… you cannot upload images, video, or PDF carousels”. ContentIn’s LinkedIn MCP server does not exist yet — it is currently in build, with a waitlist. Kalovio, the tool behind this site, sits in this group and handles text, one image, two-to-nine image swipeable posts, and PDF or document carousels.

Which one question splits the whole market?

This one: can it publish a post?

It sounds too simple to be a useful filter. It is not, because publishing is the capability that requires official permission. Reading LinkedIn can be imitated by borrowing a browser session. Writing to LinkedIn, in a way LinkedIn recognises and records, cannot — it needs a real OAuth grant with w_member_social attached, which means the tool had to be registered as a LinkedIn app and you had to approve it on LinkedIn’s own screen.

So “can it post” is a proxy for a much bigger question about how the tool is built. Ask it first and most of the market answers “no” and rules itself out in one step.

How do I check any LinkedIn plugin in a minute?

You do not need to read source code or trial the product. Three questions settle almost every case, and a tool that cannot answer them has told you something by not answering.

• Which scope does it use? A tool that posts through the official API will say w_member_social without hesitating. Vague phrasing like “official integration” with no named permission is a bare assertion, and it is remarkable how many product pages name no scope at all.
• Which endpoint does it post to? The answers are /v2/ugcPosts or the versioned /rest/posts. A tool that cannot name where a post goes is probably not sending one.
• OAuth, or a pasted cookie? If setup sends you to a page on linkedin.com and you approve there, it is OAuth — the grant is recorded in your LinkedIn settings and you can revoke it with one click. If setup asks you to copy a cookie out of your browser, it is a scraper, whatever the marketing says, and revoking it means changing your password or signing out everywhere.

A fourth question is useful if a tool promises engagement reporting: can it read posts it did not create? The official permission for that, r_member_social, is closed to self-serve apps. A tool showing you your full post history and its engagement is getting it some other way.

Which AI assistants can actually use these?

A plugin only helps if the assistant you use can load it, and support here is thinner than the marketing suggests.

• Claude supports custom connectors, including on the free plan. This is the smoothest path today.
• ChatGPT supports them through Developer Mode, which is available on paid plans (Plus, Pro, Team, Enterprise, Edu), on the web only, and the connector has to be switched on per conversation from the tools menu.
• Everything else — Gemini, Microsoft Copilot, Perplexity, Cursor, n8n and the rest — is a case-by-case question, and support changes often enough that we will not guess on your behalf. Check that app’s own documentation for whether it can connect to outside tools at all. Kalovio is confirmed working on Claude and ChatGPT and untested everywhere else.

What should you expect a posting tool to admit?

Good tools in group four are open about the same set of limits, because the limits come from LinkedIn rather than from the builder. Watch for whether a vendor mentions them at all:

• No engagement analytics. Likes, comments and impressions need LinkedIn’s partner-gated Community Management API. Link-click tracking is a different, smaller thing.
• Reconnection every couple of months. Member tokens last about 60 days and self-serve apps get no refresh token, so you sign in again periodically.
• No DMs, no connection requests, no lead scraping. None of that is inside w_member_social.
• Chat image attachments do not reach the tool. Neither Claude nor ChatGPT hands a file you drop into the chat over to a connector, so image posting runs through the tool’s own upload page or panel.
• Newer post formats may not be switched on. Multi-image and PDF posts use LinkedIn’s newer versioned interface, which may not be enabled for every self-serve app. Try it on your own account rather than trusting a feature list.

Kalovio is free while in beta, with no card, and pricing may change later. That is worth saying plainly too, because “free” on a product page rarely comes with a timeframe attached.

The short version

Most things sold as a LinkedIn plugin for AI assistants read LinkedIn rather than write to it, and several say so in their own documentation if you scroll far enough. Decide first whether you need reading or publishing. If it is publishing, ask for the scope, the endpoint and the sign-in method — and if a tool will not name all three, assume it belongs to one of the other three groups.