LinkedIn automation tools: what they do, which kind is safer, and how to run one
Every list of LinkedIn automation tools on the first page of Google is published by a company that sells one, and every one of them ranks its own product first. Tulina is not one of the tools on those lists, though it does run LinkedIn outreach (its own section is below, clearly marked), and this page ranks nobody. It sorts nine of the tools those lists keep naming by the thing that actually decides how they behave: where the software runs and which channels it sends on. Then it covers what LinkedIn's own pages say about all of them, how to run outreach at a pace an account survives, and how to do the same job from Claude or ChatGPT.
Every vendor fact below was read on the vendor's own website on 21 October 2026. Prices move, and two vendors' pages disagree with themselves; the table quotes the price card, not the meta description.
What is the best automation tool for LinkedIn?
The best LinkedIn automation tool is the one whose architecture and channels match the job: a cloud sequencer for LinkedIn and email together, a desktop app if you want the session on your own machine, and a scraper only when the job is data rather than outreach. No single tool wins every column.
| Tool | Runs as | Channels | Entry price on the vendor's page | Trial |
|---|---|---|---|---|
| Expandi | Cloud | LinkedIn and email | $99 a month, or $79 billed annually | 7 days |
| Dripify | Cloud | LinkedIn and email | $79 per user a month, or $59 billed annually | 7 days, no card |
| Waalaxy | Chrome extension and web app, campaigns kept running from the cloud | LinkedIn; email added on the Business plan | €19 per user a month (Pro) | 14 days |
| Dux-Soup (now sold as STACKED) | Desktop app (local) or cloud | $55 a seat a month (local), $99 (cloud) | 14 days, no card | |
| Linked Helper | Desktop app; the computer stays on | $15 a licence a month list price | 14 days, no card | |
| HeyReach | Cloud | LinkedIn; email through Instantly, Smartlead or EmailBison | $79 per sender a month, or $63 billed annually | 14 days, no card |
| PhantomBuster | Cloud | LinkedIn actions, plus scraping across other sites | $56 a month billed annually | 14 days, no card |
| La Growth Machine | Cloud | LinkedIn and email; calls and X on higher plans | €60 per identity a month (Basic) | 14 days, no card |
| lemlist | Not stated on its pages | Email; LinkedIn steps on the Multichannel plan | $109 per user a month on Multichannel, or $87 billed annually | 14 days |
Two things the table does not show. Waalaxy and La Growth Machine price in euros and the rest in dollars, so a straight comparison needs a conversion. And "per sender", "per seat" and "per identity" all mean one LinkedIn account: a team of three sending from three accounts pays three times.
What does a LinkedIn automation tool actually do?
A LinkedIn automation tool performs actions on LinkedIn as you, on a schedule: visiting profiles, sending connection requests with or without a note, sending a first message once someone accepts, following up, and often withdrawing invitations nobody answered. Most add an email step to the same sequence, so a person who does not accept on LinkedIn gets an email instead.
What they share matters more than what separates them. Every one of them acts under one real person's account and session, so everything it sends is, to LinkedIn, that person's activity. That single fact is why architecture and pace matter more here than features.
What is the safest LinkedIn automation tool?
The safest LinkedIn automation tool is the one that sends least, most slowly, and stops first. Where it runs changes what LinkedIn can observe, which is why vendors argue about it, but the restriction lands on your account whichever architecture sent the request.
| Browser extension | Desktop app | Cloud tool | |
|---|---|---|---|
| Where the session lives | Your browser, on your machine | An app on your machine, or a server you rent | The vendor's servers |
| Needs your computer on | Yes | Yes | No |
| What LinkedIn sees | Activity from your browser | Activity from your machine | Activity from wherever the vendor hosts it |
| The trade | Easy to install, and the extension sits in the page it automates | Your own IP and machine; you keep it running | Runs around the clock; you hand a vendor a live session |
| Examples above | Waalaxy (extension plus cloud) | Linked Helper, Dux-Soup local | Expandi, Dripify, HeyReach, La Growth Machine, PhantomBuster |
The vendors' own safety pages each conclude that their own architecture is the safe one: a desktop vendor ranks desktop apps safest, cloud vendors rank cloud safest. Read them as arguments, not measurements. None of them publishes how many of its users were restricted, out of how many, so no ranking of "safest tool" on the results page has a denominator behind it.
Is LinkedIn automation against LinkedIn's rules?
Yes. LinkedIn's help page on automated activity says it does not allow "third-party software or browser extensions that scrape, modify the appearance of, or automate activity on LinkedIn's website", and calls that a violation of its User Agreement. The agreement itself, in section 8.2, bars using "bots or other unauthorized automated methods to access the Services, add or download contacts, send or redirect messages".
Both are on LinkedIn's own pages: the help article on automated activity and the User Agreement. Neither makes an exception for cloud tools, desktop apps, AI agents or any other architecture.
That is the honest form of the "LinkedIn automation tool warning" people search for. The realistic consequence is not usually a ban on the first day. LinkedIn's help page on restrictions says an account that exceeds limits "or use[s] prohibited tools" may be restricted, and that most restrictions are removed automatically within a week. The risk is real, it scales with volume, and it falls on the person whose account it is, not on the vendor.
How many connection requests can you send on LinkedIn a week?
LinkedIn does not publish a number. Its help page on invitation limits says every member, Basic or Premium, is subject to limits, and that reaching one can restrict your invitations for about a week. The weekly figures quoted across tool vendors' blogs are their own estimates, not LinkedIn's.
What LinkedIn does state on that page: "If you reach the limit, your account may be temporarily restricted from sending invitations. This restriction typically lasts one week." The same page says Basic and free members "can include a personalized message to five connection requests per month", which is worth knowing before you build a sequence around a note on every invitation.
Because the limit is unpublished and varies by account, a fixed target is the wrong design. A budget that starts low and moves on the account's own evidence is the right one, which is what the next section does by hand.
How to automate LinkedIn outreach safely, step by step
This works with any tool in the table, or with none. It is the routine Tulina's own LinkedIn outreach process follows, written so you can follow it without us.
- Write the copy before anything sends. A connection note, a first message and one follow-up per person, stored next to their profile URL. The tool sends what is stored and never improvises.
- Start at a small daily budget. For example five invitations a day, split across several sending slots rather than one burst.
- Check the account before every slot. If LinkedIn asks for a verification code, shows a captcha or warns about a limit, stop everything and resume by hand only.
- Climb one step a week, on evidence. Move up, for example to 8, 12, 15 and then 20 a day, only after five weekdays at the current level with no warning and an acceptance rate that holds on invitations at least a week old.
- Step down on any limit warning, and hold the lower level for a week before testing again.
- Re-read each profile just before inviting. Drop people who left the company, skip people already connected, and never send a second invitation to someone with one pending.
- Stop the sequence on a reply, on any channel. A reply by email pauses LinkedIn, and a reply on LinkedIn pauses the email.
- Withdraw invitations left pending for about three weeks, a few at a time. A long tail of pending invitations is one of the things that holds an account back.
Most tools in the table can do steps 1, 2, 7 and 8. Steps 3 to 6 are judgment: what a person does when they look at the account each morning, and the part a fixed daily setting does not do.
How do you automate LinkedIn from Claude or ChatGPT?
With Tulina, Claude or ChatGPT runs the routine above on your own LinkedIn account through an MCP connector: it reads the account's status, the inbox and your connections, sends invitations and messages from your session, and writes every step into tables you can read. A person still writes the copy, sets the ladder, and lifts any pause.
The session runs on a hosted browser that Unipile, Tulina's LinkedIn provider, operates for you; there is no cookie to paste and no extension. That makes it a cloud architecture in the table's terms, and LinkedIn's rules above apply to it exactly as they apply to every other row. Paste this into Claude or ChatGPT once the LinkedIn connector is connected:
Run one LinkedIn outreach slot from my own account, at a safe pace.
1. Check my LinkedIn account status. If the session is not alive, or my
governor table (one row: status, level, daily caps, history) says
paused, stop and tell me why. Send nothing.
2. Once a day, in the first slot: set today's level from the governor's
ladder (5, 8, 12, 15, 20 invitations a day). Step down one level after
any limit error; step up one only after 5 weekdays at this level, no
blocked run in 7 days, and acceptance of at least 20% on invitations
sent 7 to 21 days ago. Write the reason into the governor's history.
3. Read my LinkedIn inbox and my newest connections. Mark anyone whose
last message is from them as replied, judging on the sender id, and
pause their lead in my email sequence. Mark new connections connected.
4. Send due follow-ups, then first messages to new connections, from the
copy stored in my outreach queue, within this slot's share of the cap.
5. Count today's sent invitations. For queued people, read each profile
first, skip anyone who left, is connected or already invited, then
send the invitation with the stored note.
6. Withdraw a few of my invitations pending for more than 21 days.
7. Never two LinkedIn writes in a row, never retry a failed write, and on
any checkpoint, captcha or restriction set the governor to paused.
Report what you sent, to whom, and what you skipped and why.
The process behind that prompt, with every step and the tools it uses:
The process in Tulina
Run LinkedIn outreach at a safe pace
Trigger · Schedule
Daily schedule in short slots
Several small runs spread across each weekday share the day's budget, so nothing goes out in one burst.
Claude runs the rest autonomously
- 1
AI checks the account before any send.check
LinkedInLinkedIn
The session has to answer alive, and nothing is written after a pause, a checkpoint or an earlier block the same day.
- 2
AI sets the day's budget from account health.
Once a day the level moves down on any warning, up one step only when every health test passes, and stops everything on a restriction.
- 3
AI records replies and acceptances.
LinkedInLinkedIn
Anyone who answered on either channel is marked replied and paused in the email sequence, and new connections are stamped with their date.
LemlistLemlist
- 4
AI sends the due messages.
LinkedInLinkedIn
Follow-ups that are due go first, then first messages to the people who accepted, each one re-read just before it leaves.
- 5
AI invites after a fresh profile check.
LinkedInLinkedIn
Each queued profile is read again first, so someone who left is dropped, a connection already made skips the invitation, and a pending one is never sent twice.
- 6
AI withdraws stale pending invitations.
LinkedInLinkedIn
Invitations left pending for weeks are withdrawn a few at a time, which keeps the backlog small enough for the budget to climb.
- 7
AI posts the weekly recap.
SlackSlack
Once a week one message gives invitations, acceptance, messages, replies by name, the current level and the backlog left.
Two things in it came from running it rather than from reading about it. LinkedIn's "is sender" flag on a message has been seen false on every thread of an account, including the account's own messages, so a reply is judged on who sent the last message instead. And a profile lookup can come back as someone else, because LinkedIn redirects old profile URLs; the identifier returned has to match the one asked for before anyone is invited.
Outreach also has a reply side, where a prospect answers and nobody does. LinkedIn outreach with Claude or ChatGPT covers drafting the messages and catching every reply.
What LinkedIn automation can't do
- Make the activity allowed. No architecture, AI agent included, changes what section 8.2 of the User Agreement says.
- Find emails or phone numbers a profile does not show. That is an enrichment tool's job, and a separate cost.
- Fix a weak list or weak copy. Automation sends a bad message to more people faster; acceptance rate is the first number to watch, and it is the one that should set your pace.
- Run several people's outreach from one account. Every invitation leaves under one person's name. A team needs one account per sender, and every tool above bills that way.
- Guarantee a number. LinkedIn publishes no weekly limit, so any tool promising a safe figure is quoting its own estimate.
Is there a free LinkedIn automation tool?
Every one of the nine tools above advertises a free trial on its pricing page, from 7 to 14 days. Whether a permanent free plan exists, and what it allows, varies by vendor and changes often, so check the pricing page on the day. Workflow builders such as n8n or Make come up for the same search; whatever they use to act on LinkedIn, it is automated activity under the same User Agreement.
The free option that is entirely within the rules is the manual one: LinkedIn's own interface, a note on the few invitations that matter, and a spreadsheet of who you contacted and when. For a few invitations a day it is often enough, and the steps above work just as well by hand. For the rest of the LinkedIn connector and what it can read and send, see LinkedIn for Claude and ChatGPT, and for the full process, run LinkedIn outreach at a safe pace.