LinkedIn Scraper MCP: Pull Profile and Post Data in 2026

A linkedin scraper mcplets your AI assistant pull LinkedIn profile and post data as a tool call instead of a scraping script. You ask Claude Code for the people who commented on a competitor's post, and it returns a structured list in the same conversation — no browser extension, no CSV export, no cookie to babysit.

The catch is that "LinkedIn scraper MCP" describes three very different architectures with very different odds of getting your account restricted. This guide covers all three, then walks through connecting SyncGTM's MCP server end to end and turning what you pull into contactable leads.

What Is a LinkedIn Scraper MCP?

A LinkedIn scraper MCP is a server that implements the Model Context Protocol and exposes LinkedIn data extraction as named, typed tools. MCP is an open standard from Anthropic that gives AI clients a common way to discover and call external services.

The practical difference is where the data lands. A traditional scraper writes a CSV that you then upload somewhere; an MCP tool returns JSON directly into the model's context, so the assistant can filter, rank, and act on it without a round trip.

If you are choosing between standalone scraping tools and an MCP-based approach, our roundup of the top LinkedIn scrapers in 2026 covers the non-MCP options in detail.

Three Kinds of LinkedIn Scraper MCP (and Their Risk Profiles)

Search any MCP registry for "LinkedIn scraper" and you will get dozens of servers that share a label but not an architecture. The difference determines whether your personal LinkedIn account is exposed.

1. Browser automation

These wrap Playwright, Patchright, or Selenium and sign in to LinkedIn as you, usually via a stored li_at session cookie. The MCP server launches a browser, navigates to the target URL, and parses the rendered page.

• Upside: Sees whatever your account can see, including first-degree connection detail and Sales Navigator views.
• Downside: Your account carries the risk. LinkedIn's automation detection can throttle, restrict, or permanently ban it.
• Downside: Brittle. A LinkedIn DOM change breaks the parser until someone ships a fix.

2. Hosted data API

These MCP servers are thin wrappers over a provider's own infrastructure. You authenticate to the provider, not to LinkedIn, and the provider returns public profile and page data from its own pipeline.

• Upside: No LinkedIn credential, no session cookie, no personal account to restrict.
• Upside: Stable schemas and predictable per-call pricing.
• Downside: Public data only. You will not get private connection notes or anything gated behind your own network.

SyncGTM's MCP server sits in this category — the same server that powers the data enrichment MCP workflow.

3. Official LinkedIn APIs

LinkedIn's Marketing Developer Platform and Talent Solutions APIs are the only fully sanctioned route. They are also approval-gated, scoped to your own pages and ad accounts, and useless for prospecting third-party profiles.

Prerequisites

The SyncGTM MCP server uses browser-based OAuth, so the list is short. Nothing here requires a LinkedIn credential.

• A SyncGTM account — free to create, and you must be signed in to it in your default browser, because that is where the authorization page opens.
• Credits on the account — LinkedIn reads start at 0.2 credits per call. See pricing for plan credit allowances.
• An MCP client with streamable HTTP support — Claude Code, Claude Desktop, ChatGPT, Cursor, Codex, Gemini CLI, or Manus.
• No API key — if a client prompts you for a token or an Authorization header, leave it blank. You have picked the wrong transport.

Step-by-Step: Connect the SyncGTM LinkedIn MCP

Setup takes about two minutes. The example below uses Claude Code — every other client points at the same server URL, https://api.syncgtm.com/mcp, with the transport set to streamable HTTP.

Per-client instructions for Claude Desktop, ChatGPT, Cursor, Codex, Gemini CLI, and Manus live in the SyncGTM MCP setup docs.

Step 1: Add the server

# Project scope
claude mcp add --transport http syncgtm https://api.syncgtm.com/mcp

# Or make it available in every project
claude mcp add --scope user --transport http syncgtm https://api.syncgtm.com/mcp

The --transport http flag matters. Pick stdio by mistake and the client will ask for credentials that do not exist.

Step 2: Authenticate in the browser

Restart Claude Code, then run /mcp. Select syncgtm, choose Authenticate, and approve the request on the SyncGTM page that opens in your default browser.

# Inside Claude Code
/mcp

# syncgtm → Authenticate → approve in browser
# Status should return to "connected"

Claude Code stores the session and reconnects on its own after that. Re-run /mcp any time to check status.

Step 3: Verify with a free call

Do not spend credits proving the connection works. The check_credits tool is free.

# Prompt to Claude Code:
"What's my SyncGTM credit balance?"

# → calls check_credits (free)

Step 4: Pull your first profile

# Prompt to Claude Code:
"Enrich linkedin.com/in/williamhgates and summarize their
current role, company, and last three positions."

# → calls linkedin_profile_enrich (1 credit)

You get work experience, education, and skills back as structured JSON. Claude Code summarizes it inline, and the raw payload stays available for follow-up tool calls.

The LinkedIn Tools You Get

Eight tools cover LinkedIn specifically. All of them are prefixed mcp__syncgtm__ in the client, and all of them accept either a full URL or a bare username or company slug.

The two directions are easy to mix up. linkedin_post_engagers answers "who reacted to this post?"; linkedin_profile_reactionsanswers "what has this person been reacting to?"

The same split applies to comments. Use linkedin_post_commenters when the post is the anchor, and linkedin_profile_comments when the person is.

Three Workflows to Run First

A single tool call is a lookup. Chained calls are where an MCP setup beats a scraper, because the assistant carries results from one step into the next without an export.

1. Warm leads from a competitor's post

# Prompt to Claude Code:
"Pull the commenters and reactors on this LinkedIn post: <post URL>.
Keep only people with Head of / VP / Director in their title at
companies with 50+ employees. Enrich each one and find their work
email. Give me a table with name, title, company, email."

# Chain: linkedin_post_commenters + linkedin_post_engagers
#        → linkedin_profile_enrich → find_work_email

Filtering happens in the model, not in a spreadsheet afterwards. That matters for cost — you only pay the 1-credit enrich and 1-credit email lookup on people who survive the filter.

We wrote up the manual version of this play in how to get warm leads from LinkedIn posts. The MCP version collapses it into one prompt.

2. Pre-call research on a single prospect

# Prompt to Claude Code:
"For linkedin.com/in/<prospect>: enrich the profile, pull their
last 10 posts, and show what they've reacted to in the past month.
Then draft three opener lines grounded in what they actually said."

# Chain: linkedin_profile_enrich → linkedin_profile_posts
#        → linkedin_profile_reactions (posted_limit: "month")

Reactions are the underused signal here. What someone quietly upvotes is often a better read on their priorities than what they publish.

3. Buying committees from hiring signals

# Prompt to Claude Code:
"Find RevOps job openings posted in the last 30 days at US companies
with 200-2000 employees. For each hiring company, pull their last
5 LinkedIn page posts so I can see what they're announcing."

# Chain: search_linkedin_job_openings (published_at: "r2592000",
#        company_size_min: 200, company_size_max: 2000)
#        → linkedin_page_posts

A company hiring for RevOps is a company rebuilding its GTM stack. Pairing the opening with the company's own posts gives you the "why now" that a job title alone never does.

For more chained patterns across other data sources, see the B2B database MCP guide.

Turn Scraped Profiles Into Contactable Leads

A scraped LinkedIn profile is not a lead. It is a name, a title, and a company — with no way to reach the person outside LinkedIn's own message limits.

The gap closes with two more tool calls on the same result set. Both live on the same MCP server, so there is no second integration to configure.

• find_work_email (1 credit) — takes the LinkedIn URL you already have and runs waterfall enrichment across multiple providers until one returns a business email.
• verify_email (0.3 credits) — checks deliverability before the address enters a sequence.
• find_personal_email (3 credits) — fallback when the work email comes back empty.
• find_mobile_number (15 credits) — the expensive one. Reserve it for accounts you have already qualified.

Order matters for cost control. Enrich, then find the work email, then verify — and only reach for personal email or mobile on the records that survive.

# Prompt to Claude Code:
"For the 20 profiles above, find work emails, verify each one,
and drop anyone whose email comes back undeliverable or catch-all.
Export the survivors as CSV."

# Chain: find_work_email → verify_email

From there the list can go straight into a sequencer or a CRM. If you are already running the HubSpot MCP alongside Claude Code, the same session can write the enriched records into your CRM without an intermediate file.

For the wider picture on running this continuously rather than ad hoc, see our guide to automated LinkedIn prospecting.

LinkedIn Terms of Service: What Is Actually Allowed

Be clear-eyed about this rather than optimistic. LinkedIn's User Agreement explicitly prohibits using bots, scrapers, or other automated methods to access the service or copy profile data. No MCP server changes that.

The legal picture is more nuanced than the ToS alone suggests. In hiQ Labs v. LinkedIn, the Ninth Circuit held in April 2022 that scraping publicly available profiles does not trigger liability under the Computer Fraud and Abuse Act, because the CFAA's "without authorization" language does not reach public websites.

That was not a win for scrapers. The case ended in December 2022 with a $500,000 consent judgment against hiQ and a permanent injunction — on breach-of-contract and state tort grounds, not the CFAA.

GDPR applies separately if you touch EU or UK personal data. You need a lawful basis — legitimate interest is the usual one for B2B prospecting — plus a documented balancing test, a privacy notice, and a working route to honour deletion requests.

None of this is legal advice. If you are scraping at meaningful volume or across regulated markets, have counsel review your process before you scale it.

Tips and Best Practices

Cap results before you run anything wide

Per-result tools bill per result. Set max_results or max_posts explicitly rather than trusting the default, especially on search_linkedin_job_openings, where 100 results is 30 credits.

Filter before you enrich

Commenter and engager lists are cheap at 0.3 credits. Profile enrichment and email lookup are 1 credit each, so let the model discard off-ICP people first.

Check the balance before batches

check_credits is free and takes a second. Running out mid-batch leaves you with a half-enriched list and no clean record of where it stopped.

Save the workflow as a project instruction

If you run the same chain weekly, write it into a Claude Code project file or a custom agent. The SyncGTM MCP repo has ready-made agent definitions you can adapt instead of starting from a blank prompt.

Expect blanks, and plan for them

Public data is uneven. Some profiles have no recent posts, some people have locked down reactions, and some work emails simply are not findable — build the fallback into the prompt rather than treating an empty field as a failure.

Conclusion

The value of a LinkedIn scraper MCP is not the scraping. It is that extraction, filtering, and enrichment happen in one conversation instead of three tools and two CSV exports.

Pick the architecture deliberately. Browser-automation servers see more but put your own LinkedIn account on the line; hosted-API servers see public data only and leave your account untouched.

Start narrow. Connect the server, run check_credits, pull one profile, then chain a commenters call into find_work_email on a post you already know performed well.

Once that loop works, SyncGTM scales it — same tools, same credits, no per-seat licence and no scraper to maintain.