LinkedIn automation in 2026: how it works

Disclosure first: gtm-api.com publishes this page and sells one of the options on it, so that option is described first. It is a LinkedIn API and MCP server that runs LinkedIn accounts you own. Your code calls the REST API, an AI agent calls the same actions as typed MCP tools, and the work runs in the cloud with nothing installed on your computer.

• Accounts you own. You sign in to LinkedIn yourself, inside the account's own cloud browser, so the password never reaches the platform.
• One cloud browser and one proxy per account. Each account keeps its own browser profile and its own proxy, so no two accounts share either, and none of it runs in the Chrome you use every day.
• Smart limits on the server. Warm-up and health-aware daily budgets are checked server-side before every action. A new account starts low and ramps up, and work over budget is parked until the budget refreshes, then resumes on its own.
• Preview, then confirm. Sends and bulk runs come back as a preview and wait for an explicit confirmation before anything reaches LinkedIn.

Pricing is per connected account, from $10/account, and API calls are not metered. The free plan runs one account with no card and no end date.

Where the software runs decides what has to stay switched on and where the LinkedIn session lives. There are four places. Our own product sits in the first row, and the other examples come from each vendor's own pages, read on 14 October 2026.

A cloud tool or an API keeps working with your laptop closed and keeps LinkedIn out of the browser you use every day. A desktop app or an extension keeps the session on hardware you control, and it stops when that hardware sleeps. The line between the two moves: Waalaxy ran as a Chrome extension until July 2026, when it went cloud only and took the extension off the Chrome Web Store.

Most tools charge per LinkedIn account per month, under one of several names: a seat, a sender, a licence or a connected account. On gtm-api.com it is a connected account, from $10/account, with API calls not metered. Elsewhere, read on 14 October 2026: Linked Helper's desktop licence starts at $15 a month, HeyReach charges $79 per sender, Expandi $99 per seat, and Unipile's API starts at €49 a month for up to 10 accounts. Data APIs that connect no account of yours charge per credit: Captain Data lists €59.99 per 1,000 credits.

Most vendors offer a trial of 7 or 14 days. A few run a free plan with no end date, and gtm-api.com has one: one connected account, no card. A trial runs out, and a free plan stays with lower caps, so read the caps before you plan around either.

Two sets of limits apply to every automated account. The first is LinkedIn's, and LinkedIn publishes few of them as numbers. It says every member is subject to a weekly invitation limit without giving the number, gives free accounts a monthly allowance of searching and browsing that it calls the commercial use limit, and caps a network at 30,000 first-degree connections.

The invitation figures people work with are observed: roughly 100 a week on free and Premium accounts and 150-200 on Sales Navigator, counted over a rolling 7-day window. On a warmed account that comes to 15-20 a day on free and Premium, or 30-40 on Sales Navigator, across about five sending days, and 5-10 a day for a new or long-idle account.

The second set is the tool's own daily budget, per account and per action. Three questions tell you what it is worth: who sets it, whether a new account starts lower, and whether it is checked on the server before each send. On gtm-api.com the platform sets it and checks it before every call. At full warm-up an account gets 25 connection requests and 50 messages a day, calls of one kind leave one at a time with a gap between them, and a new account starts lower and earns its budget by working.

LinkedIn's User Agreement, in section 8.2, tells members not to "use bots or other automated methods to access the Services". No tool changes that sentence, so automating a member account always puts the account at some risk, whichever tool does it. LinkedIn also publishes how it enforces the invitation limit: reaching it can restrict sending invitations, typically for one week, and withdrawing pending invitations does not lift the restriction.

What a tool does control is how hard it drives the account. The pages below cover that in practice: one session per account, sends paced under the limits above, a review step before anything goes out, a full stop at the first refusal, and the way back if an account is restricted anyway.

Choose on what a vendor publishes, before you connect an account. These are the questions worth answering from its own pages:

• Where it runs. One of the four places above, stated in writing.
• Whose account. An account you own and sign in to yourself, or one the vendor supplies.
• Limits, and where they are checked. A daily budget per action, a lower start for a new account, enforced on the server before each send.
• Controls in writing. Randomized timing, working hours, a proxy per account, one session per account and a review step before anything sends.
• The price unit. Per account, per seat, per credit or per workspace, and what the trial or the free plan caps.
• An API and an MCP server. Whether your code or an AI agent can drive the same account, and whether a send waits for a confirmation.

LinkedIn's own API does not open sending to outside apps: its Messages and Invitations APIs sit in LinkedIn's closed Compliance program, and no published permission lets an app send for a member. Code and agents automate LinkedIn through an account-based API instead: you connect an account you own, and the API drives it. An MCP server hands the same actions to an agent as typed tools.

On gtm-api.com the REST API and the MCP server are one contract: each of the server's 250+ typed tools maps to one endpoint, and the same server-side limits sit under both. A send from an agent comes back as a preview first, and nothing reaches LinkedIn until it is confirmed. Clients sign in through the browser with OAuth, or pass an API key as a bearer token.

Read on the dates shown. LinkedIn's pages are closed to automated re-reads, so the LinkedIn quotes are carried over from this site's earlier reads, as dated.

• gtm-api.com, developer documentation: Smart limits and warmup, MCP server, Antidetect browsers and proxies, Connect a LinkedIn account and Billing and plans. Read 15 October 2026.
• gtm-api.com, pricing. Read 15 October 2026.
• gtm-api.com, LinkedIn connection limits, the page that owns the observed invitation figures. Read 15 October 2026.
• LinkedIn, User Agreement, section 8.2, as quoted across this site.
• LinkedIn Help: Invitation limit reached, Types of restrictions for sending invitations and Commercial use limit, as checked for the limits calculator on 14 October 2026.
• Linked API, safety (a cloud browser per account). Read 14 October 2026.
• Unipile, developer protection (a fixed IP per account) and API pricing. Read 14 October 2026.
• Expandi, Activity Schedule Settings (cloud-based) and pricing. Read 14 October 2026.
• HeyReach, pricing (per sender, a proxy per sender). Read 14 October 2026.
• Dripify, pricing (campaigns run on its servers). Read 14 October 2026.
• Waalaxy, Waalaxy is now fully cloud-based. Read 14 October 2026.
• Linked Helper, pricing (desktop software, where the automation runs). Read 14 October 2026.
• Dux-Soup, pricing (the local and cloud versions). Read 14 October 2026.
• Octopus CRM, homepage (a Chrome extension). Read 14 October 2026.
• Captain Data, pricing (credits). Read 14 October 2026.