LinkedIn Outreach sends connection requests and messages from your own LinkedIn account, in a browser window Kolvera opens on your computer. It is off until you turn it on, and everything on this page is about the guard rails that sit between the queue and your account. If you read one section, read How your account is protected.
What it does, in one screen
Saving someone with the Chrome extension adds their record to Kolvera. To queue outreach, enable its draft-and-queue option, select people in the app and choose Add to LinkedIn queue, or run a campaign with a LinkedIn step. Drafts are written for you and stay editable before sending. Kolvera puts them in order and — if you have turned auto-send on — opens a LinkedIn window on this computer and works through the queue at a human pace, inside your working hours, from your own account.
Two lists, side by side — Prospects (people you sell to) and Candidates (people you place) queue separately, so a busy candidate week never eats your business-development sending.
Kolvera writes, you keep the pen — review each note before it goes, whether drafted for you or written by you. You can have it reworded, push someone to the front, hold them back, or take them out entirely.
Nothing sends until you turn it on — with auto-send off, configured sources can still build the queue when Auto-fill is on, and you can read every note on the LinkedIn Outreach page, but no invite goes out. Sending starts only once you switch auto-send on in the Chrome extension.
It runs on this computer — the sending window belongs to your browser, on your machine, signed in as you. Kolvera never holds your LinkedIn password and never signs in on a server somewhere.
Find the queue at LinkedIn Outreach in the sidebar. The switch itself lives in the Chrome extension's Settings tab, under Let Kolvera send on LinkedIn for you.
In LinkedIn Outreach → Send settings, choose LinkedIn account experience: Fresh for little recent outreach, or Established if you already send regularly on LinkedIn. Save with your daily and weekly targets. Established starts higher, but acceptance monitoring can still hold the allowance below your target. The current daily allowance is shown below the settings. Changing experience does not restart warm-up, clear restrictions or turn sending on.
How your account is protected
This is the part that matters. LinkedIn does not publish its limits and does not warn you before it acts, so every rail below is deliberately set inside what LinkedIn is thought to tolerate, and every one of them can only ever slow sending down — never speed it up. You can set lower sending targets and choose your account experience; the safeguards still apply.
Daily and weekly caps — at most 20 connection requests, 30 messages and 80 profile views a day, and 100 requests a week. These are Kolvera's maximum allowances; your personal targets and account safeguards can lower them. Weekly limits use the last seven days. When a cap is reached the queue waits for capacity — nothing is dropped and nothing is rushed.
Pending ceiling — LinkedIn limits how many unanswered requests you can have out at once, and a pile of them quietly drags your acceptance rate down. Kolvera pauses new requests at 400 outstanding and withdraws the oldest ones (below), so you never get near the hard limit. You will see this on the LinkedIn Outreach page as "Waiting on a reply".
The governor — Kolvera watches what share of your last 100 requests were accepted. Above 40% it sends at the full rate; between 30% and 40% it holds where it is rather than speeding up; below 30% it cuts the daily number back on its own; below 20% it stops sending requests altogether until acceptance recovers. Until 25 requests have gone out there is no rate worth reading, so it goes slowly by default. You will see this on the page as "How it is landing". A low rate is a message about who you are contacting and what the note says — it is never something to turn back up.
Warm-up on a fresh account — a LinkedIn account that has never automated anything does not start at the full caps. Kolvera steps it up over the first weeks, from a handful a day to the full number, and tells you which week you are in. An established account starts higher, but still not at the top.
Checkpoints stop everything — if LinkedIn shows a security check, a captcha or an unusual-activity prompt, Kolvera halts on the spot and will not send again for 72 hours. It does not retry, it does not try to click through the check, and it does not wait quietly: the halt and its reason appear on the LinkedIn Outreach page and in the extension.
A restriction is treated as worse — if LinkedIn actually restricts the account, sending stops for 7 days and warm-up starts again from the beginning afterwards. An account coming off a restriction is the most fragile it will ever be, so it is treated like a brand-new one.
Refusals are not retried — when LinkedIn itself declines an action, Kolvera counts it. After 2 refusals in a day it halves the day's caps; after 3 it stops for the day. The refused person is not tried again that day by any other route.
Old requests are withdrawn — a request with no answer after 21 days is withdrawn automatically, which frees pending headroom and protects your acceptance rate. Withdrawals have their own small daily allowance and never eat into your requests.
One second attempt, no sooner — a withdrawn request becomes eligible again on day 42, once, with a fresh note. LinkedIn enforces a cooling-off period after a withdrawal, so re-inviting sooner simply fails and spends a request for nothing. If the second attempt is also ignored, that person is done: Kolvera gives up rather than keep asking.
The same note twice is a bug, not a strategy — a second attempt always goes out with a fresh note, never the one that was ignored, and after the same wording has already gone to more than 5 people within 30 days, Kolvera stops using it. Templated-looking outreach is what gets accounts reported.
Human pace, human hours — the note is typed rather than pasted, with pauses between actions and a varying gap between people, and nothing is sent outside the working window on your campaign. There is no burst mode.
It sends what you approved, or it sends nothing — the note on the screen is checked against the note about to go, character for character. If they differ the send is abandoned, not corrected.
Never from a window you cannot see — sending happens in a real, visible browser window on your own computer. There is no hidden tab and no background session, so you can watch it and close it.
Never someone you already know — before Kolvera goes looking for a Connect button it reads the profile: if you are already connected, or an invite to that person is already showing as pending, it stops there and sends nothing. Kolvera checks its own records first, so most of these never open a profile at all. Someone you connected with by hand, years before Kolvera saw them, is caught on the page instead.
Never twice to the same person — one live invitation per person, whichever list, campaign or saved search they arrived from. Two hot lists that overlap, or a profile saved from the sidebar that a campaign was already going to reach, produce one invitation between them, not two.
An interrupted send is read, not repeated — if the browser window closes part-way through, Kolvera re-opens the profile and looks at it rather than assuming. Pending means the invitation did go out, and it is recorded as sent. A Connect button still sitting there means it did not, and the person goes back in the queue untouched. Only when the profile will not say either way does the row stop, and it says so in plain words.
You are told when LinkedIn stops you — a security check or a restriction raises a notification in your browser, once, saying what happened and what Kolvera has stopped doing. One stop is told once however long it lasts and however it changes shape; the only thing that speaks a second time is a genuine escalation, a security check that becomes a restriction. Those two are the only things this feature will ever interrupt you for, because they are the only ones you have to go and sort out on LinkedIn yourself, against a clock. There is no notification per invitation, no daily summary, and none for a pause Kolvera decided on by itself — that one needs nothing from you and is on the Kolvera icon and in the extension panel when you next look. The first time you switch this on your browser asks whether Kolvera may show you notifications; if you say no, or later turn them off for Kolvera, nothing else changes and every stop still shows on the icon, in the panel and on this page. You can turn it off yourself at Tell me if sending stops in the extension's Settings tab.
Off by default, off any time — auto-send starts off, turning it on asks you to confirm, and turning it off is immediate and never asks. Switching it off leaves the queue intact for you to send by hand.
What auto-send actually does on your computer
Auto-send is not a server sending on your behalf. It is the Kolvera extension driving a LinkedIn window on the machine in front of you, which is why the window matters as much as the caps do. Every figure below is read from the same schedule the extension runs on.
Kolvera opens the window, not you — when there is something to send, the extension opens a LinkedIn window of its own, behind whatever you are working in, and drives that. You never open a tab, paste a profile link or click Send. Close the window whenever you like; it opens a fresh one next time.
That window must stay visible — if it is minimised, or completely covered by another window, Kolvera will not type into it. The person stays queued and it tries again on a later check. Nothing is lost and nothing is rushed to catch up. This is the same rule as "never from a window you cannot see", and it is the most common reason a queue looks stuck: the window is there, but buried.
It tells you when the window is the problem, and gives you the way back — the Kolvera panel in your browser says Paused: the Kolvera window is not visible rather than sitting silent, and a Show the sending window button beside that line restores it and brings it to the front for you. You never have to go hunting through your taskbar for a window you did not open yourself. Pressing it re-reads the state straight away, so a button that worked never leaves the old sentence on screen. Leaving that window uncovered on a second screen, or simply behind your work rather than minimised, is enough; it never wants your attention or your keyboard.
One invite per check, at irregular intervals — the extension looks at the queue every 6 to 18 minutes and sends at most one invite each time. Where two actions do fall in the same visit they sit 4 to 11 minutes apart. There is no fixed start time, no hourly rhythm and no burst to clear a backlog.
It takes whole days off — about 8% of days are skipped entirely, chosen for your account so two accounts do not rest together. Nobody sends invitations every single working day, and a queue that does is the easiest kind of automation to spot. On a rest day the queue simply sits; the Kolvera panel in your browser tells you today is one of them, so an idle queue is never a mystery.
It steps aside for you — while your own window is on LinkedIn, Kolvera does not act. Two sets of hands on one account at once is exactly the pattern LinkedIn looks for.
It only runs in your sending window — the hours, the days and the timezone come from the campaign the person is enrolled in, not from a setting in the extension. Outside those hours nothing goes out, and the panel says so.
What Kolvera will never do
Raise a cap on its own — the rails only ever tighten. Nothing in the product argues its way to a bigger number.
Retry an action LinkedIn refused — a refusal is an answer, not a temporary error.
Send from a tab you cannot see — no hidden window, no headless session, no server-side login.
Send a note that differs from the one shown — if the text has changed underneath, the send is abandoned.
Message someone on your do-not-contact or unsubscribed list — the same suppression that governs email governs LinkedIn.
LinkedIn's terms, stated plainly
LinkedIn's User Agreement does not permit automated activity. Using this feature is against those terms, and the account at risk is your account, not Kolvera's — the consequence of LinkedIn objecting is a restriction on the profile you built, in the network you rely on.
That is precisely why the feature is off by default, why turning it on asks you to confirm in plain words, and why every rail above is set well inside what LinkedIn is believed to tolerate rather than close to it. We are not going to tell you this is risk-free, because it is not. What we can tell you is that Kolvera is built to be the most conservative thing running on your account: it will stop for a security check, stop for a restriction, stop when acceptance falls, and stop whenever you tell it to. Plenty of recruiters weigh this up and leave auto-send off — Kolvera still builds the queue and writes the notes, and they work the list on LinkedIn themselves. That is where every account starts, and it is the default for a reason.
Managing the queue
The LinkedIn Outreach page separates Prospects and Candidates. Each has an Invitations / Direct messages selector. Direct messages show Queued, Sent and Skipped or failed; invitation lifecycle tabs apply only to invitations. The header carries your account state, today's and this week's numbers, how many invites are waiting on a reply, and how your requests are landing.
Queued — waiting to go, in the order shown. Each row shows the date it is due.
Waiting on a reply — sent, no answer yet. The row shows how long it has been pending and the date it will be withdrawn.
Withdrawn — pulled back after going unanswered, with the date it becomes eligible for a second attempt.
Re-queued — the second attempt, with a fresh note, and the date it is due.
Given up — two attempts, no answer. Nothing further is sent to this person.
Skipped or failed — you took them out, or the send did not happen. The row says which, in plain words. People Kolvera stopped on because you are already connected, or already had an invitation out to them, land here too — nothing went wrong, there was simply nothing left to send.
No LinkedIn URL — a campaign wanted to reach them on LinkedIn but has no profile address on file, so nothing could be queued. Adding the URL to their People record does not bring them back on its own — the campaign has already moved past that step. Re-enrol them, or add them to the hot list again.
Add or remove people
In People, Desk Prospects, Talent Pool or a Hot List, select people and choose Add to LinkedIn queue. Choose Invitation or Direct message, review each recipient and draft, and exclude anyone you do not want to contact. Kolvera writes the invitation note and follow-up message from saved context. The follow-up waits for recorded connection acceptance; sending an invitation alone does not release it. Direct-message enrolment writes only a DM and requires an existing connection. Review and edit both drafts before adding people. People without a LinkedIn URL on their saved profile are omitted with a clear reason. Draft preparation runs in the background, so you can close the review and return to it. Any credit cost is shown before paid generation; queuing is free and does not turn sending on.
Use Remove on a queued row to cancel that send. Their person record remains saved. This does not withdraw invitations or delete messages already sent. If a previous delivery is uncertain, check the LinkedIn conversation and ask Kolvera support to reconcile the earlier attempt. Removing or rewriting a later draft does not clear this block.
Today: what the sender has actually been doing
Auto-send is deliberately quiet — it wakes at odd intervals, sends at most one invitation, steps aside while you are on LinkedIn yourself and takes whole days off. Every one of those is correct, and from the outside every one of them looks the same: a queue with people in it and nothing happening. Today sits at the top of the LinkedIn Outreach page and is the answer to "is this thing on?".
It is named for the day it is showing — the heading says which day you are reading, not the word "Today" over last Friday. The picker beside it lists the recent days that have something to show, labelled in plain words — Today, Yesterday, then the day and date.
The heartbeat, in one line — directly under the heading, one sentence says which of four things is true: you have sending switched off, LinkedIn or a cooling-off period has paused it, nothing has ever checked in, or how long ago the last check was. A shut laptop and a paused account both produce silence, and this is the line that tells them apart. If nothing has checked in for about 3 times the usual interval, the line says so rather than quietly showing an old time.
What is left today, and roughly how long the list will take — a pace line gives this account's remaining invitations for the day and an estimate in working days for the queue. The estimate allows for the rest days, so it is a calendar answer rather than a division.
Every check, in order, in your own timezone — each line carries the time, what happened, and who it happened to, in the same words the extension uses. Times are the account's own timezone, never the server's clock.
Quiet checks are grouped, not listed — a night of "nothing due" would write a line every few minutes and bury the day's one real send. A run of identical checks collapses into a single line with the span it covers and a count, so what actually happened stays readable.
Before anything has happened it explains itself — a brand-new account gets a sentence about what will happen and how often, instead of an empty box.
It is kept for 30 days — long enough to answer "what happened last week", and then it is cleared. It is a record of what Kolvera did on your account, not a permanent log.
Next to go, and the rows that need a look
Every other panel on the page reports what already happened. Next to go is the one place a bad note can be caught before a recipient catches it: the next 5 people in the send order, each with the exact words that would go out.
The person, where it goes, and the whole note — the name, the job title and company, whether it is a connection invitation, a direct message or a withdrawal, and on which surface. The note is shown in full and never shortened; the sentence that goes wrong is rarely in the first line.
Read from the queue, not from your browser — it is the same whether or not the sending window is open, so you can check your copy from a phone on the way in with the laptop shut since Friday.
Three things get flagged — a note still showing a variable in brackets instead of a value; a note that has already gone to too many people, with the numbers and a reminder that Kolvera stops using it past 5 in 30 days; and a step that is meant to carry a note and has none. Each flagged row says what is wrong in a sentence and links straight to that person's file.
The count is a filter, not an alarm — the line counting the flagged rows is a button, and pressing it narrows the list to the flagged rows, including flagged people further down the order than the ones on show. It never interrupts you; it is a list to work through when there is a minute.
It says how far it checked — the check reads the next 500 queued rows. If your queue is longer than that, the panel says how far down it reached and that anything past there has not been checked. "Nothing needs a look" is only ever said about rows actually read — a false all-clear over a fault further down would be worse than saying nothing.
Direct messages waiting for an accepted connection appear separately from the next sends. Sending an invitation does not release the DM; acceptance does. Their Send now control stays unavailable, and Prioritise when ready changes their place after they become eligible.
Before adding outreach and again before sending, Kolvera checks recorded outgoing LinkedIn history for that person and account, including older messages and duplicate saved profiles. Exact repeats and closely matching text require review. Check Activity and replace the draft. This comparison does not recognise every paraphrase or messages that have never been synced.
On any queued row you can:
Send now — asks Kolvera to send to that person straight away rather than waiting out the ordinary gap. See below for what it does and does not override.
Send next — pins that person to the front of the queue. Pinning changes the order, never the caps: a pinned person still waits for a free slot.
Later — pushes them to the back, without removing them. Back in order undoes it.
Skip — takes them out of the queue for good. It asks first, and the row moves to Skipped or failed so you can see what you removed.
Sending one now, and what each state means
Send now on a queued row moves that person to the front and asks the sender to pick them up on its very next look, rather than in the ordinary gap between checks. It also ignores a rest day, because a whole day off is part of the jitter and you have just said you want this one.
It overrides the wait, and nothing else — your sending window, the daily and weekly caps, the pending ceiling, a cooling-off period and the rule about people you are already connected to all still apply. If one of them stops the send, the row says which one, in the same words the rest of the page uses for it.
The row talks back while it runs — Asked to send this one now means the request is made and the sender has up to about a minute to notice it. Opening LinkedIn means it has taken the job and the window is coming up; nothing has been sent at that point. After that the row shows the outcome: sent, or the reason it was not.
One at a time — while a request is running, every Send now button is switched off. Otherwise a second press starts a second person, which is the opposite of what pressing it again means.
It says no before you press, not after — when the control cannot work — outside your sending window, an account halted by LinkedIn, or a request already in flight — the button is disabled and the reason is written beside it. You learn the rule before pressing rather than after.
It gives up rather than spin — if nothing has come back after 3 minutes, the request ends and the person goes back to their place in the ordinary order. Usually that means the browser was closed or the sending window shut in the meantime. Nothing is lost, and nothing has been sent twice.
Nothing here touches LinkedIn — the request is a note on your account that your own browser reads and acts on. The sending is still your browser, on your machine, under every rail above.
Editing a note and setting the split are not buttons on the page yet — ask the Kolvera assistant to reword a queued note or to change the balance. An edited note is still checked, character for character, against what goes to LinkedIn.
How the day's sending is divided between Prospects and Candidates is a split you control — set it to send mostly business development, mostly candidate outreach, or an even balance. The split decides the share of the same daily allowance; it never adds to it.
After a halt, once the cooling-off period has passed, a Resume sending button appears on the page header. Pressing it before the clock runs out will politely refuse and tell you when it can start — the wait is a floor, not a suggestion.
Where people come from
Saved from a LinkedIn search — the
Chrome extension saves people from a search or Sales Navigator list. Saving alone does not queue outreach. Selecting auto-draft adds invitations after drafting succeeds.
A campaign with a LinkedIn step — build the sequence as usual in
Campaigns; anyone enrolled who reaches a LinkedIn step joins the queue instead of getting an email at that step.
The nightly refill — lists tied to a Desk or a Reverse Market Profile top themselves up overnight from live
hiring signals, so an active LinkedIn campaign does not quietly run dry.
No LinkedIn URL — someone with no profile address on file cannot be queued. They are listed in their own tab rather than silently dropped, so you can add the URL and then re-enrol them or add them to the hot list again.
Everyone in the queue is a normal People record. Unenrolling someone from a campaign pauses anything still queued for them. Removing them from a hot list does not touch rows already queued.
Auto-fill, send targets and statistics
Auto-fill queues controls new automatic additions from configured campaign steps, extension auto-drafts and eligible re-invitations. Turn it off to populate Invitations and Direct messages manually from People, a Desk, Talent Pool or a hot list. Existing items remain queued, and automatic withdrawal maintenance continues. Auto-send controls delivery separately.
Under Send targets and candidate / prospect split, set daily and rolling seven-day targets independently for invitations and messages. Messages include DMs and InMail. Zero pauses that action; a blank weekly DM target means no extra weekly target. Targets cannot raise the account safeguards or warm-up allowance.
The candidate percentage applies to both invitations and messages. Eligible prioritised rows go first, and an empty stream lends spare capacity to the other. Messages waiting for connection acceptance and future scheduled items do not take the next eligible send slot.
Outreach statistics shows confirmed invitations and messages sent today, this week, this month, this year and over the tracked lifetime. Periods use your account timezone, with Monday as the start of the calendar week. These totals cover sends tracked by Kolvera, not your entire LinkedIn history. Queued, failed and uncertain attempts are excluded.
When something stops
Stopping is the design working, not a fault. Each kind of stop looks different and each one tells you what will happen next.
A halt — LinkedIn showed a security check or restricted the account. The page header says so and names the reason, the extension shows the same thing, and the cooling-off period is stated with the date it ends. Do nothing automated on the account in the meantime; use LinkedIn normally by hand. When the clock runs out, press Resume sending.
A cooling-off period — the wait after a halt. Resume will refuse until it ends and will tell you when that is. There is no way to shorten it, in the product or by asking us.
A governor pause — too few of your requests are being accepted. Sending slows or stops on its own and starts again when acceptance recovers. The fix is not a setting: it is who you are contacting and what the note says. Tighten the list, rewrite the opening line, and let the rate climb back.
An empty queue — nothing is wrong. Add saved people with Add to LinkedIn queue, select auto-draft when saving search results, enrol people into a campaign with a LinkedIn step, or check whether the people you expected are sitting in the No LinkedIn URL tab.
Auto-send switched off — existing queue items stay in place. New notes can still be prepared when Auto-fill is on; delivery stops. Turn it back on whenever you like, or work the list by hand.
If a stop does not explain itself on the page, that is a bug worth telling us about — every halt is meant to arrive with its reason attached.