LinkedIn API in 2026: The Developer Guide (Official, Third-Party & MCP)

Start with what the term means, because two different things answer to it. A LinkedIn API is any programmatic way to read LinkedIn data or act on an account without a person clicking through the interface. That covers LinkedIn's own official API and a third-party API, a managed provider acting on an account you connect. The gap between them decides what you can build.

The direct answer, before the detail: the official API does not allow outreach automation. You can't send connection requests or member-to-member messages, and profile and search access is restricted to approved partners. To search, connect, message and enrich programmatically, developers use a third-party LinkedIn API. The rest of this page covers which kinds exist, what each one costs, and what each does to the risk of losing the account you connect.

This guide is the honest map developers actually need: which official APIs exist (and which need partner approval), what only a third-party API can do, what each costs, and how to stay compliant. Facts on the official side are drawn from LinkedIn's own developer documentation.

Does LinkedIn have an official API?

Yes, but a restricted one. LinkedIn offers official APIs for sign-in, sharing, advertising, Sales Navigator, talent and learning. Two are self-serve. The valuable ones require partner-program approval. Critically, none of them send connection requests, send member-to-member messages, or search arbitrary member profiles, the exact actions outreach needs.

Self-serve (open to any developer)

In the LinkedIn Developer Portal you can use these without approval: Sign In with LinkedIn (OpenID Connect, the member's name, profile photo and email) and Share on LinkedIn (w_member_social, post, comment and like as the authenticated member on their own feed). Useful for auth and publishing. Nothing to do with reaching other people.

Partner-only (approval required)

The rest are gated behind an application and review: the Marketing Developer Platform (ads and audiences), the Sales Navigator Application Platform (SNAP) (analytics, display, and matched profile data, not open search), Talent solutions (Recruiter System Connect, Apply Connect, job posting), and Learning. Compliance APIs are closed to new requests entirely. Approval can take weeks and is limited to qualifying companies.

Here's the full official lineup, grouped by how you get access. "Self-serve" means you can switch it on in the Developer Portal today. "Partner approval" means an application and a review.

Access works the same way for all of them: you create an app in the LinkedIn Developer Portal, and the two self-serve products are usable straight away. Marketing, SNAP, Talent and Learning each need an application to the relevant partner program, and approval is limited to qualifying companies.

Notice what isn't on the list. There's no "People API", no "Messaging API", and no "Connections API" for members. Even SNAP, the closest thing to a data API, returns matched profiles to approved sales partners.

The capability map: official vs third-party

The split is consistent once you go action by action: anything about your own presence has an official route, and anything that reaches another member does not. In the table below, "official" means LinkedIn's own developer program, and "third-party API" means a managed provider like gtm-api.com acting on an account you own.

† Posting to a company or organization Page, rather than your own member feed, needs the organization scope (w_organization_social) and admin rights on that Page.

In short: the official API is built for your own presence and ads. A third-party API is built for reaching and researching other people. Almost every "LinkedIn automation" or "AI SDR" product is built on the second because the first simply doesn't expose those actions.

Why LinkedIn restricts its API

LinkedIn keeps outreach and search off the official API deliberately. Open programmatic messaging and bulk profile access would flood members and erode trust, and member data is the platform's core asset. This is a product decision backed by the Terms of Service, so building a roadmap around some future official outreach API is wishful thinking.

The stance has been tested in court, though only narrowly. The long-running hiQ Labs v. LinkedIn case concerned scraping logged-out public profiles under the CFAA, and even there LinkedIn ultimately prevailed on breach of contract, with hiQ paying $500K, deleting its data and shutting down. Authenticated, account-based automation is a separate Terms-of-Service question that hiQ leaves open. It's useful background, and it's a long way from legal cover.

Which reframes the decision. The official API can't do what an outreach product needs, so the practical question is how to do it with the least risk.

What a third-party LinkedIn API can do

A third-party LinkedIn API turns the actions a person does by hand, search, connect, message, engage, enrich, into typed endpoints an app or AI agent can call. What any given provider can actually do depends on where it sits in the market, and the category splits four ways:

• The official API. Sign-in and publishing, plus the partner-gated ads, sales and talent programs above. No outreach, no open search.
• Scrapers (Bright Data, Apify). Fetch public pages at volume. No authenticated actions, and the page is all you get.
• Dataset providers (People Data Labs, Coresignal). Sell a pre-built database of profiles. Broad coverage, but the records are as fresh as the last crawl.
• Account-based APIs (gtm-api.com, Unipile, Linked API, OutX, LinkdAPI). Act on a LinkedIn account you connect, so they return live data and perform actions. This is the only bucket that can send an invite or a message.

gtm-api.com sits in that last bucket. It groups 160+ tools into typed LinkedIn toolsets (representative below), all behind one key and one schema, so the same contract powers a REST integration and a Claude or ChatGPT agent.

LinkedIn is the live channel today. Email (Gmail, Outlook, IMAP), messengers (WhatsApp, Telegram, Instagram DMs) and calendars (Google, Microsoft) are on the roadmap on the same typed contract. Details on the agent interface are on the LinkedIn MCP server page.

Is a third-party LinkedIn API safe & compliant?

Honest answer: LinkedIn's Terms of Service don't permit third-party automation, so any tool in this space carries inherent risk. What separates providers is how much they reduce it, and the levers are the same everywhere: accounts you actually own, isolated browser sessions, and daily limits enforced before an action goes out.

Concretely, that is five mechanisms, and it's worth asking any provider which of them they actually run:

• Accounts you own. You connect them and you can disconnect them. There is no shared pool of borrowed sessions.
• An isolated anti-detect browser per account, so one account's fingerprint is never reused for another.
• A dedicated residential proxy per account. Shared datacenter IP ranges are what get a whole fleet flagged at once.
• Health-aware daily limits enforced server-side, checked before an action goes out, while refusing it still costs nothing.
• Preview then confirm available on outward actions, with an activity and audit log behind it.

GTM API reports 20,000+ accounts at under a 1% ban rate on that setup. Frédéric D., co-founder of MirrorProfiles, calls the stack behind GTM API "the safest and most reliable option we've tested".

Durability matters as much as ban-rate, and the category has a worked example. LinkedIn filed suit against Proxycurl, then one of the most-used LinkedIn data APIs, in January 2025, and the company announced its shutdown that July. Founder Steven Goh put it plainly in the shutdown post: "Regardless of the merits of LinkedIn's lawsuit, there is no winning in fighting this." Every product built on that API went dark. Because gtm-api.com is account-based and you own the connected accounts, there is no shared scraped-data pool for a single lawsuit to switch off.

One objection deserves a direct answer, because it comes up in every comparison: isn't cloud automation riskier than a browser extension on your own machine, given the traffic leaves a datacenter? It is, for tools that route accounts through a shared datacenter IP pool. That's the failure mode being described, and it's an implementation choice, which is why the second and third items on that list are per-account and residential. A session on this setup reads as one person on one home connection. The trade the other way is that an extension inherits whatever your own browser looks like and enforces its ceilings in a page script, which nothing stops you from closing mid-run.

It's also a very different risk profile from a scraper hitting pages with a borrowed cookie, and the data stays in your workspace either way. None of this is legal advice, but it's the framing a technical buyer should use. The full method is on the safe LinkedIn automation page, and the per-account ceilings are on the LinkedIn connection limits page.

LinkedIn API pricing

Third-party LinkedIn APIs price one of two ways: per connected account (a flat monthly fee, unlimited actions) or per call (metered credits per request). gtm-api.com uses the per-account model, from $19/mo per account, scaling to $5 at volume, with unlimited outreach and action calls.

There are no per-action fees on that plan, so an agent that sends thousands of messages costs the same as one that sends ten. Enrichment and signal calls are the exception, those are metered credits on top.

Per-account pricing is usually cheaper for outreach (steady sending) and per-call is cheaper for occasional lookups. Work out roughly how many actions a month you need before you compare quotes, because the two models cross over fast. The full brackets, the metered add-on tiers and the costs providers tend to bury are on the LinkedIn API pricing page.

How to get started

Getting programmatic LinkedIn access takes minutes, with no partner review to wait on: create a key, connect an account, and make your first call over REST or MCP. Both interfaces hit the same typed contract, so moving between them later changes nothing about the calls you've already written. There's a 7-day trial with no card.

1. Create an API key. Sign up and generate it in the dashboard. The 7-day trial takes no card.
2. Connect a LinkedIn account you own. It gets its own isolated browser session and its own residential proxy, and a new account warms up before it scales.
3. Make a call. Hit the REST endpoint directly, or point Claude, Cursor or ChatGPT at the MCP server and let the agent call the same tools.
4. Watch account health. Snapshots, quota-hit logs and block logs are endpoints too, so an agent can read its own headroom and back off before a restriction tells it to.

Make your first call (REST)

Every action is a typed endpoint. Sending a message is the same send_linkedin_message operation you'd call from an agent, over HTTPS:

# full reference: docs.gtm-api.com
curl -X POST https://app.gtm-api.com/linkedin/v4/api/linkedin-messages/send \
  -H "Authorization: Bearer $GTM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "linkedin_account_sid": "ln_ac_...", "ln_id": "ACoAA...", "text": "Thanks for connecting!" }'

Every parameter, guard and error code for this call is on the developer docs. The linkedin-api-examples repo on GitHub has the same call as a runnable script, in curl and TypeScript, next to nine more covering search, enrichment and connection requests.

Or skip the code (MCP)

The same tools are also exposed over the Model Context Protocol, so an agent running in Claude, Cursor or ChatGPT can call them with no integration code. The agent works from typed tool definitions, which is what lets it pick send_linkedin_connection_request on its own and still sit inside the same server-side limits a REST call gets. That's the short version. Setup, client configs and the full toolset are on the LinkedIn MCP server page.

Frequently Asked Questions

Michael SalamatovGTM engineerRuns the GTM API blog. Covers LinkedIn limits, account health and the API side of outbound.in

Last updated: July 2026 · Official-API details verified against LinkedIn's developer documentation at time of publication

Sources & Further Reading

• LinkedIn / Microsoft Learn, Getting Access to LinkedIn APIs (official products & permissions)
• LinkedIn, Developer Solutions · Become a SNAP partner
• Frédéric D., co-founder of MirrorProfiles, on the stack behind GTM API: "the safest and most reliable option we've tested"
• Ninth Circuit, hiQ Labs v. LinkedIn, No. 17-16783 (18 April 2022, on remand: the injunction against LinkedIn stands, CFAA "without authorization" likely does not reach public profiles)
• Proskauer, on the district court ruling of 4 November 2022 (N.D. Cal., 17-cv-03301-EMC: the User Agreement bars scraping, LinkedIn won on the fake accounts hiQ's contractors used, and hiQ's waiver defence was left for trial. The parties settled a month later on a $500K consent judgment)
• Steven Goh (Proxycurl / Nubela), "Proxycurl Shuts Down. Thank you." (July 2025, the shutdown announcement and the January 2025 lawsuit)
• GTM API, developer documentation · linkedin-api-examples (GitHub)
• Related: LinkedIn API pricing · LinkedIn MCP server · Safe LinkedIn automation · LinkedIn connection limits