Updated
Does LinkedIn Have an Official MCP Server?
No. As of November 2026, LinkedIn publishes no MCP server of its own, and none of its developer APIs is exposed as one. We checked LinkedIn's developer documentation, the official MCP Registry, and the main MCP directories, and found nothing from LinkedIn or Microsoft. Every "LinkedIn MCP server" you can install today comes from a third party, including the one Flux (fluxgraph.app) runs.
Updated November 2026. This is an absence-of-evidence finding, so we re-check it on every refresh.
What you'll learn
- Why there is no official server, and why that is unlikely to change soon
- The three kinds of third-party LinkedIn MCP servers, and how each one connects
- The questions to ask before you connect any of them to Claude
- What Flux's server does and does not do
Why LinkedIn has no MCP server
LinkedIn's public APIs are narrow by design. The Marketing and Community Management APIs cover your own company pages, ad accounts, and posting on your own behalf, and they require an approved developer app. There is no official API for reading other people's posts, profiles, or engagement, which is what most people mean when they ask for a "LinkedIn MCP." LinkedIn's User Agreement also restricts automated access and data collection, so the company has little incentive to publish a tool that makes automated reading easier.
Microsoft has shipped MCP servers for other products, so the protocol itself is not the obstacle. The data policy is.
The three kinds of third-party LinkedIn MCP servers
Every server you will find falls into one of three designs. The design decides what it can read and what it risks.
| Design | How it connects | What it can read | What it risks |
|---|---|---|---|
| Logged-in session (most open-source servers) | Automates a browser or reuses your LinkedIn cookie | Anything you can see when logged in | Runs under your account, so LinkedIn's automation rules apply to you |
| Official API with OAuth | Your own LinkedIn developer app and token | Your own profile, pages, and posting | Needs app approval; cannot read other people's data |
| Hosted, read-only (Flux and a few others) | A remote endpoint you connect to; the provider reads public data on its own systems | Public profiles and company pages | Depends on the provider's data source; nothing runs in your account |
The most-starred open-source option, stickerdaniel/linkedin-mcp-server, is the first kind. Its own README says to use it in accordance with LinkedIn's User Agreement and that automated access may lead to account restrictions. The most common question in its Reddit thread is whether it uses your session cookie. It does.
What to check before you install one
Four questions sort the field quickly.
- Does it use a Chrome extension or your browser session? If yes, it runs as you.
- Does it need your LinkedIn password or cookie? If yes, anything it does is attributed to your account.
- Can it take actions on LinkedIn? Liking, commenting, connecting, and messaging are where automation bans happen.
- Where does the data come from? Official API, your session, or a third-party provider. Each has different limits.
Flux answers no, no, no, and third-party provider. It reads public LinkedIn data through a data provider, not through an official LinkedIn API, and it is not affiliated with or endorsed by LinkedIn. It cannot see impressions, profile viewers, or anything behind a login.
What the Flux LinkedIn MCP server does
Flux's LinkedIn MCP server is hosted at https://www.fluxgraph.app/mcp. Connect it to Claude, Claude Code, Cursor, or any MCP client and ask about the posts and engagement of any public profile or company page. It exposes more than 60 tools: posting-pattern audits (which days, formats, and lengths beat an account's own baseline), post search, engagement trends, and who engages with whom.
A free preview covers up to 50 recent posts from a public profile or company page, within a daily budget. Full post history, weekly refresh, trends over time, and the engagement graph are on the Edge plan at $15 per month as of November 2026. Setup takes about two minutes; see connecting Claude and other AI clients.
Related questions
FAQ
Will LinkedIn release an official MCP server?
Nothing public suggests it. LinkedIn's APIs stay limited to your own content and ad accounts, and its User Agreement restricts automated reading. An official server would most likely cover posting and company-page management, not reading other people's data, which is what most LinkedIn MCP searches are looking for.
Is there a LinkedIn MCP in the official MCP Registry?
Not from LinkedIn. The registry lists third-party servers, including Flux's entry at app.fluxgraph/linkedin. A registry listing confirms the server exists and describes how it connects; it does not mean LinkedIn endorsed it. Check each entry's publisher and auth method before you connect.
Is using a third-party LinkedIn MCP server against LinkedIn's terms?
LinkedIn's User Agreement restricts automated access and data collection, so no third-party tool can promise LinkedIn's approval. Servers that automate your logged-in session put your own account at risk. Flux does not use your account and takes no actions on LinkedIn. This is not legal advice.
Which LinkedIn MCP server should I use?
Match the design to the job. To post or manage your own company page, use an official-API server with your own developer app. To read your own logged-in view, an open-source session server works but carries account risk. To research any public profile or company page without logging in, use a hosted read-only server such as Flux.
How to Use Flux MCP Tools With Claude or Cursor
Connect Flux to Claude, Cursor, or any MCP client and run LinkedIn analytics workflows conversationally — content analysis, engagement reports, prospecting, and draft scoring.
What Is a LinkedIn MCP Server? (2026)
A LinkedIn MCP server lets an AI assistant such as Claude call tools for LinkedIn data through the Model Context Protocol. What it is, what it can and cannot read, the three ways one can connect, and what to ask before you install one.