Features

Every surface a multi-account outreach team needs.

Built for agencies and SMEs alike. Twelve product surfaces, no add-ons — from per-product fit scoring to a rapport-building agent.

Two lenses, one product

Same surfaces below. They mean different things depending on who’s looking.

As you read each capability, here’s the lens to keep in mind.

Founder · SME

What this means for you

You probably need one or two of these. They mostly run themselves — the AI handles the rest. Read for outcomes (“does it book meetings?”), skip the multi-account bits.

Agency · sales team

What this means for you

You’ll use almost all of these. Read carefully for the multi-sender, pod-pacing, and per-product mechanics — that’s where the engineering depth lives.

01 · Orchestration

Run every LinkedIn sender from one console.

Most outreach tools treat each LinkedIn account as a separate workspace. LinkedReach treats them as a pool. Push a sequence across all of them, balance load, and watch sender health from a single sender strip at the top of the dashboard.

  • Manage as many of your team’s outreach accounts as you need — a single operator can run dozens at once
  • Per-sender daily caps, warm-up state, and session health visible at a glance
  • Workspaces for client separation, with role-based permissions for your team
  • Auto-pause on warning signals — the affected account stops, the rest keep working
Sender pool · Northwave Agency
12 of 50 active
Sender 01 · alex.morrison
Day 47 · warmed · 18/25 today
Active
Sender 02 · sam.lin
Day 22 · warming · 11/15 today
Warming
Sender 03 · rita.chen
Day 6 · ramp 5→8 · 4/8 today
Ramp
Sender 04 · pat.gonzalez
Restriction detected · auto-paused
Paused
02 · Sequences

Branch on every reply state.

Build a sequence as a chain of steps — view profile, connect, message 1, wait 3 days, message 2, escalate to InMail. Branch on accepted, no-reply, or replied. Conditional waits respect your campaign's send window down to the timezone.

  • Connection requests, follow-up messages, and InMail in one flow
  • Time-based and behaviour-based branches
  • Per-campaign schedule windows (start hour, end hour, days, timezone)
  • Templates for common ICPs — or import your own playbook
Q2 RevOps Founders sequence
5 steps
01
View profile
Warm up the lead before the connection ping
Day 0
02
Connection request · AI opener
Generated against profile + recent posts
+1h
03
Message 1 (if accepted)
Soft-pitch the differentiator
+2 days
04
Message 2 (if no reply)
Different angle, shorter ask
+5 days
05
InMail escalation
Only fires if Sales Navigator is connected
+10 days
3.7 · ICP-sorted scheduling

Daily caps spent on your strongest leads first.

A LinkedIn account gets 24–30 connection requests a day, and no more than 120–200 in a week. That weekly wall is the most expensive resource in the system — and it’s scarce. Most tools work the queue first-in-first-out. LinkedReach sorts by fit score, so the highest-potential leads enter the funnel first and the long tail waits.

Pre-flight a minimum-fit threshold and only your best-scoring prospects ever get queued. Faster results, fewer wasted slots, and a campaign that compounds because the strongest signals come back first — sharpening the model for everything downstream.

  • Scheduler ranks pending leads by ICP score within each account’s daily cap
  • Pre-flight threshold: drop below 70 fit and the lead never enters the queue
  • Per-campaign priority modes: high-ICP first, warm-leads first, or hybrid
  • Best-fit leads come back faster — the rest of the campaign learns from those replies
Today’s send queue · sender 01
30 / 30 cap allocated
1. Priya S. · ICP 94
VP Demand Gen · perfect product fit
Sending now
2. Marcus K. · ICP 91
VP Sales · high-fit secondary product
Queued #2
3. Anna T. · ICP 87
Founder · matches custom must-have criteria
Queued #3
Sara D. · ICP 41
Below threshold · held out of queue
Filtered
1.2 · Workspaces

Everyone who needs to look. Only senders on the bill.

A workspace holds your senders, campaigns, leads and inbox. You invite the people who need to work in it — the principal who monitors, the account manager who QAs the copy, the closer who handles replies — and give each of them exactly the permissions their job needs. Those seats are free. You pay for connected LinkedIn senders and nothing else.

Permissions are deny-by-default: a role is granted capabilities explicitly rather than inheriting everything the role above it can do. Isolation runs at the database, not in the UI, so a member of one workspace cannot read another’s leads even by guessing an id.

  • Invite by email, resend a pending invite, adjust a role, or revoke access
  • Deny-by-default capability gate — no role sees more than it was granted
  • Row-level isolation per workspace, enforced in the database
  • Manager and analyst seats are free — only connected senders are billable
  • Agencies stack one workspace per client, each fully isolated
Workspace · Northbeam
6 people · 8 senders billable
Dani R.
Billing, senders, everything
Owner
Tom A.
Reviews copy, approves drafts · cannot send
Manager · free
Priya N.
Read-only reporting across all campaigns
Analyst · free
jo@northbeam.io
Invited 2 days ago · not yet accepted
Pending
1.3 · Engagement campaigns

The people already talking about your market are leads.

A cold list is not the only place a lead comes from. An engagement campaign watches the posts your buyers are talking under — yours, a competitor’s, an industry voice’s — and captures the people who reacted or commented. Each one is enriched and scored against your products before it ever reaches a queue, exactly like an imported lead.

The sender participates rather than lurks. It reads posts, reacts, and comments in your configured voice — and when someone replies to one of those comments, or messages the sender off the back of one, that conversation lands in your inbox with the context of how you met attached.

  • Capture reactors and commenters from any post as scored leads
  • Enrichment fills title, company and profile before the fit score is computed
  • The agent answers people who comment — on the post and in the DM
  • Job adverts and company pages are skipped — a Page is not a person
  • Every engagement is attributed to the campaign that earned it
Engagement · “RevOps hiring”
14 captured today
Commented on Sasha L.’s post
31 engagers scanned · 9 above fit threshold
Captured
Marcus K. · ICP 88
Reacted · enriched · queued for connect
New lead
Reply to our comment
“how are you handling routing?” → agent answered
In inbox
Acme Corp · company page
Not a person — skipped
Filtered out
1.4 · InMail

Credits you already pay for, actually spent.

Most teams carry an InMail balance they never work through, because the credits sit in a seat nobody opens. LinkedReach reads the real balance on each sender and treats it as a sending surface: free open-profile InMails first, then the classic wallet, then Sales Navigator credits — each spent from the wallet that actually holds them.

A paid credit is never spent on a guess. The engine checks that a recipient can genuinely receive an InMail on that surface before it composes anything, and a send that never reached LinkedIn is not counted against your allowance.

  • Free open-profile InMail is used before any paid credit
  • Classic and Sales Navigator wallets tracked and spent separately
  • Subject lines carried properly — an InMail is not a plain message
  • 5–20 InMails per sender per day, rising with the sender’s phase
  • A reply window holds the sequence rather than stacking a follow-up on top
Sender 03 · InMail wallets
168 credits available
Open profile
Free — no credit spent
Preferred
Classic wallet
18 credits · spent on classic sends only
Available
Sales Navigator wallet
150 credits · previously idle
Now sendable
Cannot receive InMail
Checked before composing · no credit spent
Skipped
AI throughout the pod

One agent. Eight workflows. None of them named after a model.

Personalisation is the part of AI outreach buyers see first. It's a small share of what the system actually does. Every box below is a workflow shipping in production today — running on the same agent that learns from every reply your pod sends.

Lead qualification before the spend

Every lead is scored against your ICP before LinkedReach burns a daily-cap action on them. Returns a fit score, the reasoning, and a recommended action — bad-fit prospects never reach the queue.

Personalisation at scale

A 1–2 line opener referencing the lead's profile, role, and recent activity — per lead, dropped into the message via tokens. Hand-personalising 30+ leads a day per sender doesn't scale; this does.

Sequence drafting from a brief

Hand the AI your offer, ICP, and pod voice. It drafts the full 5-step sequence — with tighter variance on high-stakes steps (the InMail, the first connect note) and looser variance on low-stakes follow-ups.

Phrase-freshness retry

Every generated message is checked against prior sends from the pod. If phrasing overlaps too much, it's regenerated once with explicit "do not reuse" instructions. Keeps the pod sounding human as it scales.

Reply classification & routing

Every inbound is tagged interested / objection / not now / wrong person / negative / out-of-office — routed to the right pod member's queue, with high-intent ones surfaced first. Closers stop reading "thanks but no thanks" all day.

Agent Mode auto-reply

Drafts the response, asks qualifying questions, and proposes calendar slots. Operators choose whether replies send automatically or surface for one-click approval — new campaigns ship with approval-by-default.

High-stakes safety gating

Every inbound is keyword-scanned for legal, GDPR, contract, pricing, refund, "remove me" — the kind of phrases where an auto-reply would be a disaster. On a match, auto-send is disabled for the thread, an alert is filed, and a human picks it up.

Long-thread compaction

When a conversation runs 10+ messages, the agent keeps the topic-setting first message plus the recent nine and elides the middle. The AI stays on-topic in 30+ message threads instead of losing the plot at message twelve.

A note on what we don't say

We won't tell you which AI vendor is behind it — that detail will change a dozen times in the next two years and we'd rather you bet on the workflow than the model. The agent is a frontier LLM, swappable behind the scenes, and it gets sharper as your pod sends more replies.

03 · AI personalisation

Openers that reference what the lead actually posted.

Personalisation isn't first-name token replacement. Each opener is generated against the lead's profile, recent posts, and your ICP brief — producing a sentence that reads like it took you ten minutes to write. Then a phrase-freshness check compares it against everything the pod has already sent, and regenerates if it sounds templated.

  • Per-lead opener grounded in profile data and recent activity
  • ICP brief stored at the campaign level — same agent, your voice
  • Phrase-freshness retry: catches templated phrasing before it ships
  • Tone preview & safety filter before any message goes out
  • Fallback to your written template if the AI can't clear the quality bar
Generated opener preview
Lead · Priya S.
CONTEXT
VP Demand Gen at Cinder. Posted last week about ABM tooling fatigue and "stitching together five different point solutions."
DRAFT MESSAGE
Hi Priya — saw your post on ABM stack stitching. We're building the orchestration layer specifically to kill that problem for outreach across multiple senders. Worth a 15-min look?
Phrase-freshness check
Compared against 1,847 prior sends · 12% overlap
Cleared
04 · Agent Mode

Operator-approval and full-autonomy modes for inbound.

Most tools stop at the opener. Agent Mode keeps going. When a lead replies, the agent reads the message, scores warmth, drafts a contextual response, and offers real calendar slots once warmth crosses the threshold you set. Operators choose whether replies are sent automatically or surfaced for one-click approval. The reply-preview / approval flow is on by default for new campaigns — flip it off once you trust it.

  • Inbound reply analysed for warmth, intent, tone, length, and meeting interest
  • Drafted reply uses your tone, your ICP, and the conversation history
  • Calendar integration proposes real open slots and books on confirmation
  • Approval-by-default mode for week one · auto-send after you flip the switch
  • Every drafted, edited, and sent message logged with full audit trail
Agent Mode · Priya S.
Booked · 2m ago
INBOUND
Yeah, makes sense — happy to chat. Tuesday or Thursday work, mornings ideally.
Intent · Interested
Confidence 94% · warmth flag: hot
Auto-replied
AGENT REPLY (sent)
Tuesday morning works great. I have 9:30 or 10:30 ET open — either of those land for you?
Calendar hold · Tue 9:30 AM
Tentative until lead confirms
Booked
3.8 · Conversational sequence builder

Talk the sequence into existence. No outreach experience required.

Most outreach platforms assume you already know what a good sequence looks like. LinkedReach assumes you don’t. Tell the AI in plain English — “I want to reach VPs of marketing at B2B SaaS, my offer is X, my goal is to book a demo” — and it builds the sequence with you, suggests the tone, and ships it to the canvas in one click.

Already know what you’re doing? Hit “Generate with AI” in the canvas and the AI proposes the best-fit sequence shape for your offer. Either way: founders ship a working campaign on day one without copying someone else’s playbook.

  • Plain-English chat builds the full multi-step sequence with you
  • One-click “Generate with AI” proposes the optimal shape for your offer
  • Per-campaign tone & approach selection — direct, trigger-based, insight-challenger, social-proof, value-first, and more
  • Edit any step by hand — or tell the AI to tweak it and re-apply
Build with AI
Conversational builder
YOU
Reach heads of operations at logistics SaaS, 50–200 people. Want them on a 30-min discovery.
AI · ready to apply
5-step sequence: profile view → connect with a hook on logistics ops pain → 2-day wait → first message with a sharp insight → 4-day wait → case-study follow-up. Tone: conversational, peer-to-peer.
Tone · conversational · trigger-based
High reply rate per LinkedIn norms for this segment
Recommended
3.9 · Rapport-building agent

Builds the conversation. Then books the meeting.

Most “AI replies” are auto-responders that fire the same CTA on every reply. LinkedReach’s agent is built to build relationships. On every inbound, it scores warmth, reads the lead’s role, skills, and recent posts, and writes a reply that actually advances the conversation — matching the lead’s tone and length when configured to.

The meeting CTA only fires once warmth crosses the threshold you set. Before that, the agent keeps the conversation going. When it’s time, it pulls real open slots from your calendar (Microsoft 365 supported), proposes them, and books on confirmation. Operators choose whether replies send automatically or surface for one-click approval.

  • Warmth scored on every inbound (0–100, deterministic)
  • Tone & length matched to the lead’s reply style when configured
  • Meeting CTA gated on warmth threshold — doesn’t fire too early
  • Real-calendar slots from Microsoft 365 / Teams — not a static booking link
  • Optional reply-preview / approval flow keeps a human in the loop
Agent reading the lead · Priya S.
Rapport-building turn
INBOUND
Interesting — we’re actually reviewing options in this space. What’s your take on the build-vs-buy debate?
Warmth · 67 · warm
Curious + engaged · meeting threshold not yet met (75)
Reply, don’t pitch
DRAFTED REPLY
Honest take: build-vs-buy depends on whether the orchestration logic is your core IP or just plumbing. Saw your post on the data-platform stitching problem — usually that’s the tell. Where are you leaning?
3.6 · Per-product AI fit scoring

Paste a URL. The AI reads what you sell.

Setup used to mean a 90-minute “tell us about your ICP” form. LinkedReach replaces it with one field. Drop in the page where you describe your product or service — pricing page, landing page, even your homepage — and the AI extracts the name, the description, and the ideal-customer language in five seconds. You edit, accept, ship.

Then it scores every lead against every product you sell. The dashboard shows you which product is the best pitch for which prospect — before the first connection request goes out. Multi-product teams stop sending the wrong pitch to the right buyer.

  • Auto-extracts product name, description, and target use-case from a URL
  • Scores every lead per product, not just per ICP — surfaces the best pitch
  • Five scoring dimensions: title fit, industry, location, company size, custom criteria
  • Deterministic scoring (temperature 0) so the same lead always produces the same number
Per-product fit · Marcus K.
3 products scored
Demand-Gen Platform · 92
Title fit + posted about ABM tooling fatigue last week
Best pitch
Sales Enablement Suite · 63
Adjacent role, less direct match
Moderate
RevOps Reporting Tool · 28
Wrong buyer persona for this product
Weak
06 · Lead qualification

ICP fit and warmth, scored on every reply.

Your closer doesn't want to read every "thanks but not now" reply. LinkedReach scores every replied lead on ICP fit, warmth, and intent — routes the qualified ones to the closer queue, archives the rest, and surfaces aggregate signal across the campaign.

  • 0–100 ICP fit score against your stored brief
  • Warmth flag (cold / warm / hot) based on reply tone and intent
  • Auto-routing to the right teammate's queue
  • Aggregate signal: which titles, industries, and company sizes convert
Qualification queue · this week
12 new hot
Replies
187
Qualified
42
Hot
12
Priya S. · ICP 92
VP Demand Gen · Cinder · 200–500 employees
Hot
Marcus K. · ICP 88
VP Sales · Mainline · 50–200
Warm
Sara D. · ICP 41
Account Exec · out of ICP · routed to archive
Filtered
3.10 · Lead qualification scoring

Every prospect is graded before a single connect request lands.

ICP scoring is usually a pre-import filter. Tools tell you “these 500 leads are a good fit”, you trust the bucket, and by the time the message goes out the AI has forgotten why the lead was in there. The reasoning vanishes. The trust is in a number.

LinkedReach grades each lead against each product on five weighted dimensions per call. Title carries 30 points, industry 20, location 10, company size 10, custom criteria 30. Every lead is scored against every product you’ve defined — not once into a generic ICP bucket — and the reasoning is exposed in the lead row so an operator can audit any score.

  • 5-dimension weighted scoring (title 30 / industry 20 / location 10 / company size 10 / custom 30)
  • One score per (lead × product) — not one bucket
  • Reasoning surfaced in the UI — every score auditable
Score breakdown · Marcus K.
5 dimensions weighted
Title fit · 28 / 30
VP Demand Gen matches buyer persona
Strong
Industry · 18 / 20
B2B SaaS, mid-market
Match
Custom criteria · 26 / 30
Posted about ABM tooling fatigue
Hit
3.14 · Per-step temperature tuning

A connect note runs at lower variance than a chase-up.

Most personalisation runs at one temperature for all message types. The connect note (which you can never edit again, ever) gets the same generative variance as the day-7 follow-up. One off-tone connect request and the lead is burned permanently — the same risk profile as a low-stakes chase-up makes no sense.

LinkedReach tunes temperature per step, matched to recoverability. InMail runs at 0.6, the connect note at 0.7, regular message at 0.8, follow-up at 0.85. Higher-stakes steps — the ones you can’t take back — get tighter variance. Lower-stakes follow-ups get looser language so the pod stays fresh across thousands of sends.

  • InMail 0.6 / connect 0.7 / message 0.8 / follow-up 0.85
  • Higher stakes = tighter variance (less off-tone risk)
  • Lower stakes = looser language (more freshness across thousands of sends)
Temperature curve · sequence steps
Stakes ↓ · variance ↑
InMail · 0.6
Highest stakes · tightest variance
Tight
Connect note · 0.7
Cannot be edited after send
Tight
Follow-up · 0.85
Recoverable · freshness over caution
Loose
3.15 · High-stakes phrase gating

Some replies need a human, immediately.

An AI auto-replier is one “I’m a lawyer, you have 24 hours to delete my data” message away from a deliverability nightmare. The model doesn’t understand which replies cross from “let me handle that” into “you really do not want me to handle that”. The cost of getting that line wrong is not symmetric.

LinkedReach keyword-scans every inbound for legal, GDPR, contract, pricing, refund, “remove me”, lawyer, complaint and similar high-stakes patterns. On match: thread-level auto-send is disabled, a the agent alert is filed, and a human picks up via the inbox. No silent auto-response on a thread that should never have one.

  • Keyword scan + key_objection detection on every inbound
  • On match: thread-level auto-send disabled, Agent alert filed
  • Human picks up via inbox — auto-reply never silently fires on legal
Inbound gating · thread #4821
High-stakes phrase scan
INBOUND
Please remove me from this list and confirm under GDPR you have deleted my data.
Match · “remove me”, “GDPR”
Two high-stakes patterns hit
Auto-send disabled
Agent alert filed
Routed to inbox · human picks up
Awaiting operator
3.17 · Long-thread compaction

A 30-message thread doesn’t make the AI forget the original ask.

Standard chat-completion APIs degrade once threads get long. The model starts to overweight recent messages and forgets the original ask. Your pod opens about a May meeting, the prospect engages over 12 messages across 3 weeks, and by message 13 the AI is offering to send a calendar link to a meeting that’s already booked.

LinkedReach compacts long histories by structure, not by inference. When a conversation exceeds 10 messages, the agent keeps the very first message (sets the topic) plus the most recent 9. The dropped middle is replaced with a single “[N earlier messages elided]” marker. No extra LLM call to summarise; the topic is preserved by structure.

  • First message preserved (sets the topic, never drops)
  • Most recent 9 messages preserved (current state)
  • Middle compacted to a single marker — no extra LLM call needed
Thread compaction · lead #2117
17 messages → 10 retained
Message 1 · preserved
Topic-setting first message · never drops
Anchor
[7 earlier messages elided]
Middle compacted · no LLM summary call
Marker
Messages 11–17 · preserved
Recent 9 · current state retained
Recent
3.18 · Language

Sells in the language your market actually buys in.

Set a sender’s language and it holds all the way through. The agent searches in that language, reads posts in it, writes connection notes, messages, follow-ups and comments in it, and answers replies in it — without you keeping a parallel set of templates per market.

It is one authority, injected once, rather than an instruction bolted onto each prompt — so a sender working the German market does not quietly revert to English three steps into a sequence, and does not surface prospects it has no way of writing to.

  • One configured language drives search, reading, writing and replies
  • Posts the sender could not credibly write under are skipped, not fumbled
  • Grounding survives the language rule — a translated message keeps its facts
  • The high-stakes reply gate reads meaning in every language, not just English
  • Per-sender, so one workspace can run several markets at once
Sender 05 · DACH
Language: German
Search terms
“Vertriebsleiter”, “Personalplanung”
German
Connection note
Written in German · grounded in their post
German
Reply: “Bitte keine Werbung”
Opt-out understood · sequence stopped
Handled
Post in Portuguese
Outside this sender’s language — skipped
Skipped
07 · Safety

Built around LinkedIn’s published norms.

Hard sending caps aren’t a feature you toggle — they’re enforced at the worker level. Operators can’t override them from the UI. Every action is randomised in timing, each account’s connection is managed for you, and any LinkedIn restriction instantly pauses the affected account and tightens pacing across the rest of the pod.

  • Per-phase daily ceilings — 24–30 connection requests and 60–120 messages per account, per day
  • A hard weekly invite ceiling of 120–200 — the wall that actually paces an account
  • Minimum 60-second gap between actions, randomised above it and stretched during warm-up
  • New accounts start at 20% of their allowance and reach full volume over about five weeks
  • Each account’s connection is managed for you — no setup, nothing to install
  • Any LinkedIn restriction → account auto-paused, operator notified, pod pacing tightens
  • Per-campaign send-window enforcement (start hour, end hour, days, timezone)
Sender 01 · today
all green
Connection requests
18 of 30 daily cap · 84 of 190 this week · randomised gaps 61–112s
Within cap
Messages sent
62 of 105 daily cap
Within cap
Send window
Mon–Fri 09:00–17:00 ET · respected
Active
Account health
No restrictions · 47 days clean
Healthy
3.11 · Anti-burst pacing

Sixty seconds between actions is the floor, not the average.

LinkedIn’s anomaly model isn’t watching raw speed — it’s watching the rhythm. Tools that batch fifty connects in five minutes trigger soft warnings the same day. A bot fires actions on a stable cadence; a person pauses between conversations to think.

LinkedReach enforces a 60-second minimum gap, scaled by multipliers that stack. Warmup day 1–3 multiplies by 3. Recent timeout multiplies by another 3. A new sender that just got warned waits 9× the floor before its next action. Capped at 4 minutes so a stale state can’t pin a worker indefinitely.

  • 60-second floor × multipliers — warmup-day = 3×, recent timeout = 3×
  • Multipliers stack (warmup-day-1 + recent timeout = 9× the floor)
  • 4-minute hard cap so worker queues never starve
Gap calculation · sender 03
Warmup day 2 · recent soft warning
Floor · 60s
Base inter-action gap, every sender
Base
Warmup multiplier · 3×
Day 1–3 scaling
Stacked
Effective gap · 540s
9× floor · capped at 240s
Capped at 4 min
3.12 · Soft-warning propagation

A warning on one account slows down the rest of the pod.

Per-account pacing tools let a warned sender’s pod-mates keep firing at full volume. LinkedIn sees the unchanged behaviour across the pod and within 24 hours restricts the whole team in lockstep. One sender takes a hit; the team takes the consequences.

The moment LinkedIn flags a sender — a restriction, an “unusual activity” notice, a “please slow down” — LinkedReach catches it and pushes a 3× pacing multiplier across the whole pod for 24 hours. A second flag inside that window auto-pauses the account.

  • LinkedIn restrictions caught the moment they surface on a sender
  • First strike: pod-wide 3× pacing multiplier for 24h
  • Second strike inside 24h: account auto-paused
Pod response · warning detected
Sender 02 · soft-warning banner
Sender 02 · first strike
“please slow down” matched
3× pacing · 24h
Senders 01, 03–08
Pod-wide multiplier applied
Slowed in lockstep
Second strike · 24h
Account auto-paused on second flag
Account paused
3.16 · Cross-sender deduplication

A pod of eight senders is one pod, not eight independent ones.

When a pod has 8 senders, the import lists overlap — every agency adds the same target accounts to multiple campaigns. Without dedup, two SDRs in the same pod connect with the same VP of Marketing in the same week. The prospect thinks they’re being spammed by a content farm; the agency loses the lead twice.

LinkedReach runs workspace-scoped dedup on the (recipient_linkedin_id × action_type × 7-day window) tuple. The second sender’s queued action is silently postponed past the window, not failed. The action stays in the queue and fires once the window clears — no lead is dropped, no sender collides.

  • Workspace-scoped dedup — pod members never collide on the same target
  • (recipient × action × 7-day window) is the unique key
  • Postpone past the window, don’t fail — action stays queued
Dedup decision · queue check
Recipient × action × 7-day window
Sender 01 · connect Priya S.
Tuesday · sent successfully
Sent
Sender 04 · connect Priya S.
Thursday · same recipient, same action
Postponed past window
Sender 04 · rescheduled
Auto-fires once 7-day window clears
Queued
05 · Unified inbox

Every sender's replies in one feed.

Switching between fifty LinkedIn tabs is how operators burn out. The unified inbox surfaces every reply across every connected sender, attributed to the right account, sortable by intent, warmth, or campaign.

  • Reply-only view across the whole sender pool
  • Filter by sender, campaign, intent, or warmth flag
  • Quick-respond inline from any sender's identity
  • Conversation history threads automatically
  • Mark hot, snooze, archive, or hand off to a teammate
Unified inbox
38 unread
Priya S. · via sender 01
"Yes — happy to chat next Tuesday."
Hot
Marcus K. · via sender 04
"Send me a one-pager and I'll loop in finance."
Warm
Anna T. · via sender 02
"Not the right person — try Jordan in revops."
Referred
Jordan R. · via sender 03
"OOO until Monday."
OOO
08 · Integrations

Plug into the rest of your stack.

Native sync with the CRMs that matter. Webhook triggers for everything else. Zapier for the long tail. Lead activity, reply state, and meeting bookings flow into your pipeline without an admin running CSV exports every Friday.

  • Native HubSpot & Salesforce sync (bi-directional, contact + activity)
  • Webhooks for connection accepted, replied, qualified, meeting booked
  • Zapier and Make.com integrations for the long tail
  • A scoped REST API over every object in the product — see below
  • Calendar integrations: Google, Microsoft 365, Calendly
Connected integrations
4 active
HubSpot
Bi-directional · last sync 12s ago
Connected
Google Calendar
Agent Mode booking enabled
Connected
Webhooks · reply.qualified
POST to ops.cinder.io/leads
Active
Zapier
3 zaps running
Active
8.1 · API & assistant connectors

Ask your own AI assistant about your pipeline.

Every object in the product — senders, campaigns, sequences and their steps, leads, inbox, agents, activity — is addressable over a scoped REST API. Connect Claude or a ChatGPT Custom GPT to it and you can ask in plain language which campaigns are stalled, how many leads from the Berlin list replied, or whether the new agent is out-performing the old one, and have the answer come from real records rather than a guess.

The boundary is deliberate, and it is not the one people expect: an assistant can read and shape everything you own, and cannot send anything on LinkedIn. There is no send endpoint and no send tool. Outbound runs only inside the safety engine, which owns pacing, daily caps and the warm-up ramp — so nothing you connect can talk an account past a limit.

  • Connects over MCP for Claude, or an imported action schema for a Custom GPT
  • Read-only or read-write keys, scoped when you mint them — start read-only
  • Keys are stored hashed and shown once; revoke and re-mint at any time
  • Deletes go to a recycle bin and stay restorable for 30 days
  • Setup is one command or one URL — copy-ready snippets in the app
Developer · API keys
2 active
“Claude desktop”
read · last used 4 minutes ago
Active
“Ops GPT”
read + write · can edit, cannot send
Active
“Which campaigns are stalled?”
Answered from live records · 3 campaigns, reasons attached
Read
Send a message on LinkedIn
No such endpoint — the engine owns outbound
Refused
9 · Dry-run simulator + shareable reports

See what the AI would send. Before it sends anything.

Open the simulator on any campaign and the agent runs the full sequence against synthetic-but-realistic leads. You read every drafted message in your voice, watch the agent handle a fake angry reply, see the warmth score after each round, and confirm the meeting slots it would offer hit your actual calendar — without burning a real connection request.

When the run finishes, one click generates a public read-only URL of the entire report — every message, every warmth trajectory, every AI decision rationale. Send it to your pod lead, your RevOps person, or your manager for greenlight signoff. They don’t need a LinkedReach account to view it.

  • 1× to 5000× playback — step through every action or watch a 14-day sequence collapse into 30 seconds
  • Five reply personas: warm, curious, busy, hostile, ghosted
  • Real-calendar read-only check confirms slot-picker against your actual Microsoft / Google availability
  • Shareable URL persists indefinitely — come back to a sim run weeks later
  • Full deep dive on the Agent Mode page →
Sim Report · shared with pod lead
linkedreach.ai/simulator/share/aB3xK9_2qZw1
LEAD · SARAH (CURIOUS PERSONA)
Title: VP RevOps · Acme · Final: building_interest · Warmth: 72
PILOT WOULD DRAFT
Honest answer: depends on whether your finance team owns the supplier-onboarding tooling or just signs off. Curious which way it’s set up — that changes what I’d show first.
DECISION RATIONALE (TURN 2)
Reply intent: curious + buyer-blocked. Skipping pitch escalation, asking a qualifying question that surfaces the actual buying motion.

See it on your own senders.

Invite-only early access. Hand-reviewed within one business day. Connect a LinkedIn account, build a sequence, send your first message in under ten minutes.