How to Build a LinkedIn MCP Server With n8n (Connect Claude & AI Agents)

You can build a working LinkedIn MCP server in n8n in under 15 minutes, using n8n's native MCP Server Trigger node and LinkedIn node, no custom code required. The result is a URL that Claude, ChatGPT, or any MCP-compatible AI agent can call to publish a LinkedIn post on your behalf, mid-conversation, without you touching LinkedIn directly.

This guide walks through the exact build: what you need, the step-by-step setup, how to connect an AI client to it, and (just as important) what this specific approach can't do and how it compares to the other LinkedIn MCP routes covered elsewhere on this site. (Facts verified August 6, 2026, against n8n's own documentation.)

Key takeaways

• n8n has a native MCP Server Trigger node that turns any workflow into its own MCP server, no custom code needed.

• Connecting it to n8n's native LinkedIn node exposes a "Create a post" tool that AI agents can call directly.

• This is an API-backed, OAuth-authenticated approach, the same account-safety category as any approved LinkedIn app, not session-cookie automation.

• It's genuinely different from the n8n LinkedIn scheduler guide already on this site: that one runs on a timer, this one runs on-demand when an AI agent calls it.

• Claude Desktop and Claude Code don't speak n8n's MCP transport natively, you'll need a small bridge, covered below.

What Is an n8n MCP Server, and Why Use It for LinkedIn?

An MCP server is a set of tools an AI agent can discover and call mid-conversation. n8n's MCP Server Trigger node turns a single workflow into exactly that: you attach "tool" nodes to it (in this case, the LinkedIn node), and n8n exposes a URL that any MCP client can connect to.

Why build this yourself instead of using a hosted LinkedIn MCP server? Two reasons. First, you keep full control of the LinkedIn credential, it lives in your own n8n instance, not a third-party vendor's infrastructure. Second, you can extend it. Once the workflow exists, adding a second tool (a Slack notification after posting, a Google Sheets log, a content-approval step) is a matter of wiring in another node, not waiting on a vendor's roadmap.

The trade-off is setup time and technical comfort. This isn't a one-click install, it's a genuine build, even if a fast one.

Does n8n Have Official MCP Support?

Yes. n8n ships two purpose-built nodes for this: the MCP Server Trigger (what we're using here turns a workflow into an MCP server) and the MCP Client Tool (the reverse lets an n8n workflow call tools from an external MCP server). n8n also has a separate, broader feature called Instance-level MCP, which lets an AI client build and edit workflows across your whole n8n instance; that's a different use case from this tutorial and worth knowing about, but not what we're building today.

Per n8n's own documentation, the MCP Server Trigger node "acts as an entry point into n8n for MCP clients" by "exposing a URL that MCP clients can interact with to access n8n tools." It connects only to tool nodes, not the rest of your workflow canvas, which keeps the exposed surface intentionally narrow.

What Do You Need Before You Start?

Four things, all free or already-owned for most readers:

1. An n8n instance — n8n Cloud (fastest to start) or a self-hosted instance. Self-hosting requires a bit more setup for the LinkedIn OAuth app (covered below), but the MCP mechanics are identical either way.

2. A LinkedIn Developer App — free, created at the LinkedIn Developer Portal, with Sign In with LinkedIn and Share on LinkedIn products enabled.

3. An MCP-compatible AI client — Claude Desktop, Claude Code, or ChatGPT in developer mode all work, with one caveat covered in Step 6.

4. 15 minutes. This build genuinely doesn't take longer than that once your LinkedIn app credentials exist.

How Do You Build the LinkedIn MCP Server, Step by Step?

Step 1: Create Your LinkedIn Developer App and Credentials

In the LinkedIn Developer Portal, create a new app and request the Sign In with LinkedIn (OAuth2) and Share on LinkedIn products. This grants the w_member_social scope,  LinkedIn's own documentation describes it as required to "post, comment and like posts on behalf of an authenticated member." Note your Client ID and Client Secret; you'll need both in n8n.

In n8n, go to Credentials → New → LinkedIn OAuth2 API. If you're on n8n Cloud, you can skip manual app creation entirely and just click Connect my account to authorize through your browser. If you're self-hosting, you'll paste in the Client ID and Secret from your LinkedIn app and complete the OAuth flow manually.

Step 2: Add the MCP Server Trigger Node

Create a new, empty workflow in n8n. Add the MCP Server Trigger node as the workflow's starting point. On its setup screen, n8n generates a random MCP URL path by default (you can customize it if you want something memorable). This is the node that will become your server's entry point once the workflow is live.

Step 3: Connect the LinkedIn Node as a Tool

Add a LinkedIn node to the canvas and connect it to the MCP Server Trigger as a tool node (not a regular workflow step, the MCP Server Trigger specifically expects tool connections). Set the operation to Create a Post, choose Post As: Person (or Organization, if you've been through LinkedIn's Community Management App Review for that scope), and select your LinkedIn credential from Step 1.

Here's the part that makes this genuinely AI-agent-ready rather than just automated: leave the post text field's value to be set dynamically. n8n's LinkedIn node explicitly supports being "used to enhance the capabilities of an AI agent," meaning its parameters (the post text, in this case) can be populated automatically from what the AI agent sends, rather than hardcoded. That's what turns "a node that posts one fixed message" into "a tool an AI agent can call with any content it wants to post."

Step 4: Secure the MCP Endpoint

Back on the MCP Server Trigger node, set an authentication method. Bearer Auth is the simplest. Generate a token, save it as a credential, and keep it somewhere safe; you won't be able to view it again after saving. Skipping this step means anyone who discovers your MCP URL could call it and publish to your LinkedIn account, worth taking seriously given what's at stake.

Step 5: Activate the Workflow and Copy the Production URL

Save and activate the workflow. n8n registers a production MCP URL the moment you do — this is the URL your AI client will connect to. Copy it now; you'll need it in the next step.

Step 6: Connect Your AI Agent

Here's the one genuine wrinkle in this whole build: n8n's MCP Server Trigger communicates over SSE and streamable HTTP, not stdio. Claude Desktop and Claude Code natively expect either a local stdio process or a remote HTTP endpoint depending on configuration, and n8n's transport doesn't map cleanly onto stdio out of the box. To bridge this, most builders use a small proxy tool like mcp-remote, which translates between the two.

Once bridged (or if you're connecting a client that speaks streamable HTTP natively, like ChatGPT in developer mode via a public HTTPS endpoint), add the server the same way you would any custom MCP connector: paste in your n8n production URL, include your Bearer token in the auth header, and let the client discover the tool. From that point on, asking your AI agent to "post this to LinkedIn" should trigger the LinkedIn node and publish through your account.

Is This Safe for Your LinkedIn Account?

Yes, in the specific sense that matters most. This entire build runs through LinkedIn's own API, authenticated via OAuth2 with permissions you explicitly granted through LinkedIn's consent screen. That places it in the same account-safety category as any approved LinkedIn app, not the session-cookie, browser-automation category that carries real restriction risk under LinkedIn's User Agreement. Nothing in this workflow touches your li_at session cookie or drives a logged-in browser.

One safety gap worth naming directly, though: this build, as described, publishes immediately when called. There's no confirmation step showing you the exact text before it goes live. That's a real risk if your AI agent's context ever includes untrusted text (a pasted email, a fetched web page) alongside a live publish tool. 

If you build this yourself, consider adding a manual approval step (an n8n Wait node, a Slack approval message, or similar) between the MCP trigger and the LinkedIn node, so a human confirms the exact post text before it publishes, the same principle behind why any publish-capable MCP server should gate on confirmation, covered in more depth in our LinkedIn automation and MCP terms guide.

What Can't This Workflow Do?

Worth being honest about the limits, since the reference build this tutorial is based on genuinely ships with just one operation:

• One operation only, unless you add more. Out of the box, this exposes "Create a post" and nothing else, no reading your analytics, no listing existing posts, no scheduling. n8n's LinkedIn node supports more operations than that, so you can expose more tools by adding more node connections, but the minimal build only does the one thing.

• No built-in scheduling. This publishes immediately when the AI agent calls it. If you want scheduled, calendar-driven posting instead of on-demand AI-triggered posting, that's a different workflow, see our n8n LinkedIn post scheduler guide for that specific build.

• No voice-matching. The AI agent calling this tool writes the post text using whatever it already knows about your voice (or doesn't). This workflow has no memory of your past posts and doesn't train on your writing style the way a dedicated tool like ContentIn's AI Ghostwriter does.

• Self-hosting adds real setup overhead. If you're not on n8n Cloud, the LinkedIn OAuth2 app configuration is a manual, multi-step process, and running the MCP Server Trigger reliably at scale (multiple webhook replicas) requires routing all /mcp* traffic to a single dedicated instance, an infrastructure detail most solo builders won't hit, but worth knowing if you're deploying this for a team.

How Does This Compare to Other LinkedIn MCP Routes?

The honest positioning: this route sits between "hosted and turnkey" and "fully custom." You get more control than a vendor's hosted server, less setup than the fully open-source scraping route, and (unlike most of the alternatives) the ability to extend the workflow yourself without waiting on anyone. For the full landscape this fits into, see our comparison of LinkedIn MCP servers.

Frequently Asked Questions

Do I need to code anything to build this?
No. Every step here uses n8n's visual node editor: the MCP Server Trigger, the LinkedIn node, and the credential setup are all point-and-click. The one piece of "technical" work is bridging the transport for stdio-based clients like Claude Desktop, which means running a small proxy tool rather than writing code.

Can I use this with ChatGPT instead of Claude?
Yes, provided ChatGPT can reach your n8n instance's production MCP URL as a public HTTPS endpoint, ChatGPT's developer mode only supports remote MCP servers, so if your n8n instance isn't publicly reachable (e.g., running locally), you'd need to expose it first.

Does this cost anything?
The build itself is free using n8n's free tier for light use, plus a free LinkedIn Developer App. Costs only appear if your usage grows past n8n's free execution limits or you move to a paid n8n plan for reliability/scale.

Is this the same as n8n's LinkedIn post scheduler?
No, that's a separate, already-published guide on a timer-based mechanism (n8n + Google Sheets, firing on a schedule). This guide is about an AI agent calling LinkedIn on-demand through the MCP protocol. Different trigger, different use case.

What happens if my n8n workflow goes down?
The MCP tool simply becomes unavailable to your AI client, it won't silently fail or post something wrong, it'll just fail to connect. Worth testing the connection periodically if you're relying on this for anything beyond casual use.

The Bottom Line

Building a LinkedIn MCP server in n8n is a genuinely achievable weekend-afternoon project, not just a concept, n8n's native MCP Server Trigger and LinkedIn nodes handle the hard parts, and the whole thing runs through LinkedIn's official API rather than anything resembling session automation. 

Where it falls short of a purpose-built option is voice quality (it writes however your AI agent already writes) and publish-safety (no confirmation step unless you build one in yourself). If you want the control and extensibility of building it yourself, this is a real, working path. 

If you'd rather have voice-matching and a built-in confirmation step without building any of this by hand, that's the gap ContentIn's own LinkedIn MCP server is being built to close, though as covered elsewhere on this site, it's still pre-launch, on a waitlist, not something you can install today.