The exact 7-agent system that runs your entire outbound pipeline automatically — from buying signal to booked meeting, without touching a thing.
Reply rates tank. Emails hit spam. You hire someone new, wait two weeks, and hope the next campaign works.
The issue isn't the copy. It isn't the offer. It's that every part of the outbound process — signal detection, research, personalisation, sequencing, reply handling, booking — depends on a human doing it consistently. And humans don't scale.
So we replaced every manual step with an agent. Seven of them, each with one job, running in sequence — 24 hours a day.
ATLAS monitors every active campaign continuously. When performance drops, it doesn't wait for a weekly review.
Feed ATLAS your open rate, reply rate, and sequence data. It diagnoses whether the drop is copy fatigue, list quality, or a deliverability problem — and returns three specific changes to make this week. Not vague advice. Not a dashboard. Three actions with the reasoning behind each one.
This is the exact build order. Follow it sequentially — each phase depends on the one before it.
Minimum viable stack to start: Claude Code + n8n + Airtable + Apollo + Smartlead. Add the rest as you scale each agent.
This is the central database all 7 agents read from and write to. Every lead, signal, email, and reply is logged here.
Go to airtable.com/account → API → copy your Personal Access Token. Then open your base, click Help → API docs — your Base ID is in the URL (starts with "app...").
Create a Trigify account and set up 3 signal monitors: (1) keywords "hiring SDR OR outbound OR sales automation" on LinkedIn posts, (2) job title changes to Head of Sales or VP Sales at target companies, (3) new job posts containing "BDR" or "outbound".
Trigify sends a webhook when a signal fires. This is what wakes SCOUT up in n8n.
In n8n, create a new workflow called "SCOUT — Signal Intake". Add a Webhook node as the trigger. Copy the webhook URL and paste it into Trigify's notification settings.
After the webhook, add an HTTP Request node that calls the Claude API. Send the signal data and ask Claude to score it 1–10 for ICP fit and buying intent.
Add an IF node checking if score >= 7. If true, write the signal to the Signals table in Airtable and trigger the IRIS workflow via a webhook call. If false, log and discard.
Triggered by a webhook from SCOUT. The workflow runs 5 enrichment steps in sequence: Apollo lookup → Clay enrichment → Hunter email verification → LinkedIn post scrape → Claude pain hypothesis.
Take the unverified Apollo email and run it through Hunter to verify. If Hunter confidence is below 80%, try the pattern {first}.{last}@{domain} and verify that too. Use whichever returns valid.
Feed all enriched data to Claude and generate the 2-sentence pain hypothesis that CIPHER uses as the foundation for the email.
Use the Airtable node in n8n to create a new record in the Leads table with all enriched fields. Set status to "New". Then trigger the CIPHER workflow via webhook.
Triggered by IRIS webhook. Reads the enriched lead record from Airtable and sends the full profile to Claude Code with the CIPHER system prompt.
Use n8n's JSON parse node to extract each email object. Loop through them and create one Airtable record per email in the Sequences table. Then trigger NEXUS.
Always validate that Claude returned exactly 4 emails + 1 DM. Add an IF node to catch malformed responses and retry once before alerting Slack.
When NEXUS activates, it creates a new Smartlead campaign for the lead and uploads the 4 emails as sequence steps with the correct delay days.
In Smartlead, go to Settings → Webhooks. Add your n8n ECHO workflow webhook URL as the "Reply Received" event. This is what instantly triggers ECHO when any reply arrives.
This is the most critical step. Without this webhook, warm replies sit in your inbox unactioned. Test it by sending a test reply to yourself before going live.
Use n8n's Wait node to delay 10 days, then call the PhantomBuster API to trigger the LinkedIn Message Sender phantom with the lead's LinkedIn URL and the DM copy from CIPHER.
Triggered by the Smartlead reply webhook. Receives the full reply body, sender email, and campaign ID.
Add a Switch node in n8n with 5 branches. POSITIVE → write to Replies table + send Slack alert with draft. NEGATIVE → update Airtable status to DNC. OOO → pause campaign + schedule re-activation. REFERRAL → create new lead in Airtable and trigger IRIS.
Set up your cal.com booking page. Enable email confirmations. Add 3 questions to the booking form: current outbound process, team size, and monthly email volume. These feed directly into APEX's pre-call prep.
In Cal.com, go to Settings → Webhooks and add your n8n APEX workflow URL for the "BOOKING_CREATED" event. This triggers the pre-call prep sequence automatically on every new booking.
Set the Schedule node to run every Monday at 8:00 AM. This workflow pulls data from Airtable and Smartlead, runs it through Claude, and posts the report to Slack.
Create 3 separate n8n workflows that run every 6 hours and check thresholds. If any threshold is breached, ATLAS posts an immediate Slack alert — no waiting for Monday.
Do not go live until every checkbox passes. One broken webhook in the middle of a pipeline silently kills leads with no error. Always test with real data on a dummy lead first.
Book a 45-minute call. We'll show you exactly how we'd deploy all 7 agents for your ICP.
Book Your Call →