Clay + LinkedCamp: The Agency Signal-Based Stack

Claygent crossed one billion runs in 2024, and at this point it's the default research layer for any serious outbound team. Funding triggers, headcount deltas, tech installs, job changes — Clay catches the signal, enriches the record, and writes the opener line. That part is solved.

What Clay doesn't do is send on LinkedIn. And the SERP is flooded with Clay + HeyReach, Clay + Aimfox, Clay + Lemlist tutorials written for solo operators running one inbox. None of them address what actually breaks at agency scale: multi-client account isolation, per-seat warmup pacing, compliance with LinkedIn's tightening limits, and the unit economics of Clay credits across 10+ clients.

This post lays out the clay linkedin automation stack we see working inside agencies right now — the one where Clay handles signal + Claygent research, and LinkedCamp handles safe, multi-account LinkedIn + email sending. Step by step, with real numbers and the prompt structures that don't get flagged by 360Brew.

Why Clay alone isn't enough (and why HeyReach isn't the agency answer)

Clay is exceptional at three things: pulling intent signals, running Claygent across a list to do open-ended research, and stitching together waterfall enrichment so your coverage rate stays above 75-80%. It's a GTM engineering platform, not a sender.

The LinkedIn sending tutorials online almost universally point to HeyReach or Aimfox. Fine for one operator running one account. The problem starts when you're an agency running outbound for eight clients, each with two SDR seats, each with its own warmup state, each with its own signal feed. HeyReach's per-seat pricing and lack of true client workspace isolation forces agencies into spreadsheets and a lot of swivel-chair work.

The real agency question isn't "can I connect Clay to LinkedIn?" It's "can I run 16 LinkedIn accounts at signal-triggered pace without one restriction tanking a client retainer?"

That's the gap LinkedCamp was built for, and it's why a destacked GTM workflow usually lands on Clay + LinkedCamp rather than four point tools stitched together.

The five signals that actually move LinkedIn reply rates

Not all triggers are equal. Across the campaigns we audit, five Clay-sourced signals consistently produce LinkedIn reply rates in the 15-25% range when paired with a Claygent-written opener:

1. Funding rounds (Series A through C, 0-45 days old) — highest urgency, hiring is about to spike
2. Job change into a buying role (new VP/Director, 14-60 days in) — they need quick wins
3. Headcount growth >20% in 90 days in a target department — operational pain is real
4. Tech install/uninstall via BuiltWith or Wappalyzer — disqualifies and qualifies at once
5. Website visitor signals from RB2B or Vector — warmest possible intent

RAIN Group's research shows top performers convert 52% of qualified opportunities in just five touches. Stacking two or three of these signals on the same account is what gets you into that performance tier — single-signal campaigns rarely break 8% replies anymore.

The Clay → LinkedCamp workflow, end to end

Here's the exact wiring for a funding-triggered LinkedIn play. This is the template we hand to new agency clients on day one.

Step 1: Signal capture in Clay

Build a Clay table with a Crunchbase or Harmonic source filtered to: Series A-C, raised in last 30 days, headcount 50-500, target industries. Refresh daily. Expect 40-120 new companies per week in most SaaS verticals.

Step 2: Find the buying committee

Use Clay's people-finder waterfall (LinkedIn Sales Navigator → Apollo → Prospeo) to pull 2-4 contacts per company by title pattern. Set a coverage target of 80% — anything lower and you're sending to ghosts.

Step 3: Claygent research pass

This is the prompt that matters. Don't ask Claygent to "write a personalized opener." Ask it to extract one specific fact the prospect would recognize as real research. Prompt structure:

"Visit {{linkedin_url}} and {{company_website}}. Find one specific thing this person has posted, shipped, or hired for in the last 60 days that connects to {{funding_use_of_funds}}. Return one sentence, no preamble, no superlatives."

The output goes into a column called opener_fact. You'll use it raw — not as an AI-written sentence, but as a fact you reference in a human-written opener template.

Step 4: Push to LinkedCamp via webhook

Clay's HTTP API node fires a webhook into LinkedCamp's campaign endpoint with: prospect URL, opener_fact, funding_amount, role_context, and the assigned client workspace ID. LinkedCamp routes the prospect into the correct client account's sending queue.

Step 5: Multi-touch cadence inside LinkedCamp

The sequence we run for funding signals:

• Day 0: Connection request, no note (acceptance rates are 8-12 points higher without notes for cold)
• Day 2 after accept: Opener referencing opener_fact + funding
• Day 5: Voice note or short follow-up
• Day 9: Email touch (synced via LinkedCamp's email sender, SPF/DKIM/DMARC enforced)
• Day 14: LinkedIn breakup message

Five touches, two channels, signal-grounded. This is the cadence that produces the reply rates we benchmark against.

Why the clay linkedin automation stack lives or dies on per-account limits

LinkedIn's 100 connection requests per week cap is the binding constraint everyone underestimates. If you're running 10 client accounts at 100 requests/week, that's a 1,000-request weekly capacity ceiling — and the moment any one account spikes above 150 sends with low acceptance, the LinkedIn volume tax kicks in and that account's reach collapses.

The agency math:

• 16 seats × 80 requests/week (safe pace) = 1,280 weekly touches
• At 32% acceptance and 18% reply on accepted, that's ~74 conversations/week
• At 15% meeting conversion, that's ~11 meetings booked weekly across the agency

Those numbers only hold if you're rotating actions across accounts you own at per-account volume that respects LinkedIn's limits. LinkedCamp's account rotation, per-seat warmup curves, and randomized send windows are built specifically so one client's aggressive campaign doesn't compromise another client's seat health.

Claygent prompt design: writing openers 360Brew won't flag

LinkedIn's 360Brew model is actively deprioritizing AI-generated openers — the structural fingerprints (em-dashes, "I noticed," three-clause sentences, generic compliments) get clustered and suppressed.

The fix is to never let Claygent write the message. Let it extract the fact. Write the template yourself.

Bad Claygent prompt: "Write a personalized LinkedIn message to {{name}} about their recent Series B." Output reads as AI every time.

Good Claygent prompt: "Find the specific product or hire they announced post-funding. Return as: '{product/hire} - {one-sentence context}'. No commentary."

Then your opener template is human-written and reuses the fact:

"Saw the {opener_fact}. We help {ICP} solve {pain} during exactly that phase — worth 15 min?"

This sidesteps pattern saturation because the variable is genuinely researched, but the surrounding structure is a stable human template. Best of both.

Unit economics: what this actually costs per client

Agencies blow Clay credits in two places: running Claygent on unqualified lists, and re-running enrichment on records that haven't changed. Tight ops:

• Clay Explorer or Pro: ~$349-800/month depending on credit volume
• Claygent runs: budget 1 credit per qualified prospect, ~2,500-5,000/month for a mid-size agency
• LinkedCamp seats: per-seat pricing across all client accounts, no per-client minimum
• Email infra: SPF/DKIM/DMARC compliance is non-negotiable — Google and Yahoo reject ~84% of non-authenticated bulk mail

The agencies hitting target margins gate Claygent behind a qualified-fit filter (ICP score >70) so they're not burning research credits on prospects who won't get a touch anyway.

Where this stack breaks (and how to keep it from breaking)

Three failure modes we see repeatedly:

1. Signal staleness. Funding data older than 45 days converts at half the rate of fresh. Set a hard TTL on every Clay table.
2. Account-level over-sending. One eager campaign manager bumps a client account to 200 requests/week, acceptance tanks, and that account is in the LinkedIn doghouse for 30+ days.
3. Claygent hallucinations on thin sources. If the prospect has no recent posts and a sparse profile, Claygent will invent something plausible. Add a confidence flag and route low-confidence rows to a manual review queue.

The agencies running this stack cleanly treat it like infrastructure, not magic. Daily monitoring on acceptance rate, per-seat send volume, and reply velocity — same way you'd watch deliverability metrics on cold email.

• Clay solves signal + research + enrichment; it doesn't send on LinkedIn — and the standard HeyReach/Aimfox tutorials don't scale to multi-client agency ops
• The agency clay linkedin automation stack that works: Clay table → Claygent fact extraction → webhook to LinkedCamp → multi-account safe sending with per-seat warmup
• Five signals drive 15-25% reply rates: funding rounds, job changes, headcount spikes, tech installs, and website visitor identification
• Never let Claygent write the message — let it extract one specific fact and plug it into a stable human-written template to dodge 360Brew's AI-opener suppression
• Respect per-account limits (80-100 requests/week max), gate Claygent runs behind ICP score to control credit burn, and monitor account health daily