SDR-IN-A-BOX
Marketing Agents on Autopilot

A step-by-step guide to building AI agents that find LinkedIn leads, grab their contact info, write personal emails, and book meetings — all by themselves. Based on Greg Isenberg's Marketing Agents Masterclass.

Steps

$500

Monthly

10K

Emails/mo

THE CORE TRICK

You don't need 5 salespeople costing $22,000/month. Build ONE AI agent for $500/month. The agent finds people who already care about your topic (they raised their hand by commenting), writes a personal email, and books demos.

The Big Idea: Marketing is Code

Paid ads are like throwing money into a fire. You pay $22 per 1,000 people who might see your ad. Most never buy. But what if you could find people who already care about your topic — people who already raised their hand by engaging — and reach out to them directly?

Marketing used to be a mystery. Now it's just code — a robot employee you build once that works 24/7. Marketing agents are the new coding agents.

FOR HIGH SCHOOL STUDENTS

Think of it like a vending machine. You set it up once: put in money, choose your snack. Then it runs by itself. This AI agent is your vending machine — but instead of snacks, it finds potential customers and books them meetings.

10K

Emails Per Month

500+

Leads Found

80%

Cheaper Than Ads

Manual Work (After)

The 6-Step SDR-in-a-Box Pipeline

One agent. Six steps. Zero manual work. Here is what happens, in plain English.

01 / Find Creators

Search LinkedIn by Topic

Go to LinkedIn Search. Type your niche keyword (like "AI marketing"). Find creators who post daily about this topic. Add 10-20 of them to a list. These are your starting points.

02 / Find Hot Posts

Spot the Viral Posts

Look at each creator's recent posts. Most get no likes. A few go viral with 100x more engagement. Those posts are where your best customer prospects are.

03 / Extract Engagers

Scrape the Comments

Anyone who liked or commented on a viral post already cares about your topic. Use Apify or Ampify to scrape these people. They raised their hand. You just collect their names.

04 / Enrich

Find Their Email

Take each person's LinkedIn URL and look up their email. Try getleads.io first (cheapest). If that fails, try Apollo. If that fails, try Origami. Check all emails with Million Verifier.

05 / AI Writes Outreach

Send Personal Emails

An AI checks if each person is your ideal customer. If yes, it writes a 3-line email that references the exact post they commented on. Then sends it through Instantly.

06 / Inbox + Demo

Book the Meeting

When someone replies, your AI gets notified instantly. If "yes," it sends a Cal.com link to book a call. If "no," it thanks them and sets a 90-day re-engage list. The agent never stops following up.

Follow-Up Timeline

80% of sales close after the 5th message. Your agent never gives up and never forgets a name.

SIMPLE RULE

You send ONE message per week to each lead. Start with their name and a reference to the post they engaged with. End with a question. Do this for 7 weeks. Most will say yes.

D0NEW

Day 0 — First Contact NEW LEAD

Send your first email or LinkedIn DM. Reference the exact post they commented on: "I saw your comment on [creator]'s post about [topic]." Short and personal. No sales pitch.

Day 7 — First Follow-Up

"Just checking if you saw my note about LinkedIn engagers. Happy to hop on a 10-minute call?" Keep it light and easy.

D14

Day 14 — Value Add

Send something useful — a case study or data point. No pitch. "Thought you'd find this engaged playbook useful." Help them first, ask later.

D30

Day 30 — New Hook

"Saw [another creator] post about this too — thought of our chat." Fresh angle, same topic. They see you're staying on top of it.

D60

Day 60 — Social Proof

"We just helped [similar company] get 3x more leads using this exact method." Social proof beats sales pitches every time.

D90

Day 90 — Cold Lead Reset

Find a NEW trending post about the same topic. "Saw this trending now and thought of you." If still no reply, move to the 6-month list.

6 Months — Snowball Round 2

Every 6 months, automatically re-run the pipeline for cold leads. Find a fresh trending post. New angle. Same topic. The snowball keeps rolling.

Ads vs Agents

Paid ads cost $22 per 1,000 impressions. This AI pipeline costs $500/month flat and finds real buyers.

LinkedIn
Ads

$22K / 1M
fake bots

AI Agents

$500 / mo
real humans

WHY IT'S CHEAPER

With ads, you pay $22 every time 1,000 people see your ad — even if zero buy. With the agent, you pay $500/month flat. The agent finds people who already engaged with your content (they want to buy), then sends a personal email. This costs a fraction of ads.

Cold Email Stack

Hypertide inboxes: $100/mo
Instantly plan: $97/mo
Apify/Ampify: $30/mo
getleads.io: $50/mo
Total: ~$277/mo

LinkedIn DM Stack

HeyReach: $79/mo
LinkedIn Premium: $30/mo
Apollo: $50/mo
Origami: $40/mo
Total: ~$199/mo

Infrastructure

Railway server: $5-15/mo
ClickHouse DB: $10/mo
Cal.com: Free
Million Verifier: $20/mo
Total: ~$55/mo

TOTAL MONTHLY COST ~$500 / month vs $22,000 for ads

The Tools You Need

Each tool has an API. Your agent talks to them so you do not need to.

🔍

Search for creators in your niche. Find the people posting daily about your topic.

linkedin.com/search

⚡

Ampify

One API key gets you data from LinkedIn, Twitter, and 20+ platforms. Feed results to your agent.

API key → JSON

🎯

API Maestro

Hosts pre-built scrapers. Pick the LinkedIn engagement scraper and run it with zero coding.

No-code scrapers

🤖

Apify

Run scrapers on demand. The post-reactions actor returns everyone who liked or commented on any post.

POST_REACTIONS actor

🔗

getleads.io

First email lookup. Cheaper than Apollo. Takes a LinkedIn URL and returns an email address.

URL → email

🌌

Apollo

Backup email finder. Catches emails that getleads.io misses. Also has phone numbers.

Backup lookup

🔷

Origami

Final fallback. Runs the whole enrichment chain so you call one API instead of three.

One API → all data

✅

Million Verifier

Checks if an email is real. Bad emails get you marked as spam. This stops that.

Email validation

📱

Lead Magic

Finds phone numbers. If email fails, try texting instead.

Phone lookup

✉️

Instantly

Send cold emails at scale. Webhooks fire when someone replies — your agent gets notified instantly.

API + webhooks

💬

HeyReach

LinkedIn DM automation. Send connection requests and DMs at scale. Has webhooks too.

LinkedIn outbox

📅

Cal.com

Booking calendar. Your agent sends this link to schedule calls. Checks if meetings are confirmed.

Scheduling

How It All Connects

Your AI agents run on a server. They talk to LinkedIn, email tools, and calendars. You set it up once, then it runs by itself.

1. Cron Schedule

A timer runs everything every day at 2 AM UTC. No manual start needed.

2. Agent Server

A Python server (Railway or VPS) holds all your API keys as environment variables.

3. Data Storage

ClickHouse stores your leads, emails sent, and reply data. Everything is tracked.

4. API Integration

The server talks to LinkedIn, Instantly, Cal.com, and 7+ other tools through their APIs.

Your Claude Code Agent Prompt

Copy this prompt into Claude Code. It will build your SDR agent for you.

# PROMPT FOR CLAUDE CODE — SDR ENAGERS AGENT
# Build the 6-step marketing automation pipeline

You are a growth engineer. Build a Python script that runs a 6-step marketing agent pipeline. Do EXACTLY what is described. No skipping. Follow all steps in order.

def find_creators(niche):
    # 1. Search LinkedIn for creators in this niche
    # Return 10-20 creator profile URLs
    # Use Apify LinkedIn Scraper API
    # Cache results to creators.json

def find_trending_posts(creator_urls):
    # 2. Scrape each creator's recent posts
    # Score by engagement (likes + comments)
    # Return top 3 posts per creator
    # Cache results to trending.json

def extract_engagers(post_urls):
    # 3. For each trending post, use Apify
    # POST_REACTIONS actor to scrape all
    # people who liked or commented
    # Output: engagers.json with profiles

def enrich_contacts(engagers):
    # 4. Waterfall email lookup for each engager:
    #   a) getleads.io (cheapest)
    #   b) Apollo (if getleads fails)
    #   c) Origami (if Apollo fails)
    # Also try LeadMagic for phone numbers
    # Validate ALL emails with Million Verifier

def send_outreach(leads):
    # 5. For each valid lead:
    #   - Use LLM to check ICP fit (10-500 employees, right title)
    #   - If fit: write 3-line email referencing their post
    #   - Send via Instantly API in batches of 50
    #   - Log to sent.csv

def manage_inbox():
    # 6. Webhook receiver on /webhook/instantly
    # When someone replies:
    #   - Positive: send Cal.com link to book
    #   - Negative: thank + add to 90-day re-engage list
    #   - Question: answer + Cal.com link
    #   - Non-reply after 7 days: follow up

def main():
    # Run all 6 steps in order every day
    # Schedule via cron: daily at 02:00 UTC
    # Store all API keys as env vars
    # Use ClickHouse for data storage
    creators = find_creators("AI marketing")
    posts = find_trending_posts(creators)
    engagers = extract_engagers(posts)
    leads = enrich_contacts(engagers)
    send_outreach(leads)
    manage_inbox()
    # Loop back to step 1 after 6 months for cold leads

if __name__ == "__main__":
    main()
# End of prompt

SDR-IN-A-BOXMarketing Agents on Autopilot