Clay Workspace Optimization · Silver GTM

The Clay Credit Playbook

5 ways teams are burning credits under the new pricing — and the exact settings that stop it. Every fix grounded in Clay's own docs.

Maintained by Victor Kalu · silvergtm.com ↓ Save as PDF

First, the trap

Clay's 2026 pricing meters two things. Actions are the platform work — every enrichment, AI run, export, and sync. Data Credits are the data itself.

Actions are a hard ceiling you cannot top up. Launch gives you 15,000/mo, Growth 40,000/mo (month-to-month). Run out and your tables stop mid-campaign — the only fix is upgrading your whole tier.

Data Credits you can top up (at a 30% premium) and they roll over. But the pool shrank hard: teams who had 50,000 legacy credits now start at 6,000 on Growth.

Source: Clay — Actions & Data Credits; Plans & Billing.

Five fixes you can apply this week — not theory. These are the exact leaks burning real teams right now.

The five leaks

Stop accidental re-runs

A RevOps team adjusted one formula column and it silently re-triggered enrichments across the table — 60,000+ credits gone in a day. Clay won't refund "ran as expected," even when it's a mistake.

The fix

Turn off Auto-Update on tables you're not actively running — Run Settings → toggle Auto-Update off (or ⋯ next to the table name → Auto-Update Columns).
When you edit a formula or column, choose not to overwrite existing columns, and switch on the table setting that stops edits from re-running existing enrichments.

Saves: Actions — and the Data Credits each re-run would spend. [Clay: Credit Conservation]

Bring your own LLM key — and run a cheap model

A consultant building deep-research Claygents said they "stopped being viable" under per-credit AI pricing. Another ran GPT-4.1 nano at 0.5 credits a call and still felt the squeeze once the pool dropped to 6k.

The fix

Connect your own OpenAI / Anthropic key under Settings → Connections.
Run your AI / Claygent columns on a low-cost model — GPT-4.1 nano or GPT-4o mini. On Clay's keys a nano run is ~0.5 Data Credits; on your key it's a fraction of a cent to OpenAI and 0 Data Credits. (The Action still counts either way.)

Saves: Data Credits. [Clay: Claygent Builder; "your OpenAI key vs ours" FAQ · OpenAI pricing]

Bring your data-provider keys

If you already pay for Apollo, ZoomInfo, Clearbit, Findymail, Prospeo, or an email verifier — you're paying for that data twice the moment Clay charges credits for it.

The fix

Connect those keys under Settings → Connections. The enrichment then costs 0 Data Credits — you pay the provider directly.
The Action still counts, but the data is free to Clay.

Saves: Data Credits. [Clay: Credit Conservation; Actions & Data Credits]

Filter before you enrich

Most teams enrich a 5,000-row table when 200 rows are actually in-ICP. The cheapest enrichments are the 4,800 you never run.

The fix

Pre-filter to your real targets before enriching.
Set enrichment columns to "only run if empty" (Run Settings → Only run if) so you never re-enrich data you already have.
Enriched a company once? Pull it into other tables with a single-row Lookup instead of re-enriching.

Saves: Actions and Data Credits. [Clay: Conditional Runs; Credit Conservation]

Pick the cheaper provider

A RevOps consultancy buying mobile numbers for BDR teams: at ~10 credits a number, 100 numbers eats a sixth of a 6k pool — and some come back as bad data.

The fix

Same data point, wildly different price — and Clay shows the per-provider cost in the enrichment panel before you run.
Mobile number: 2 → 25 credits. Company funding data: 1 → 10. AI model: 0.5 → tens of credits.
Pick the 2-credit option over the 25-credit one for the same field.

Saves: Data Credits. [Clay enrichment panel; community-documented spreads]

The reference

What each operation costs

Costs shown as Actions / Data Credits per operation. 0 = free. Source: Clay docs.

"No result returned = no charge."
Operation	Actions	Data Credits
Find People / Companies (sourcing)	0	0
Marketplace enrichment (Clay's keys)	1 / record	~6–20 / record
Marketplace enrichment (your key)	1 / record	0
Claygent / AI run (Clay's keys)	1 / run	token-based
Claygent / AI run (your key)	1 / run	0
AI Formula (no LLM)	0	0
AI column	1 / row / run	token-based
CRM import (into Clay)	0	0
CRM export / sync / create	1 / record	0
CSV import / export	0	0
Outbound HTTP call	1 / call	0
Inbound webhook	0	0
Lookup — single row	0	0
Lookup — multiple	1 / run	0
Send data to table	0	0¹
Clay Sequencer (lead added)	1 / lead	0
External sequencer export	1 / record	0
Signals (intent / hiring / funding)	1 / record	maybe
Ads audience export	1 / record	0
Auto-Update re-run	1 / record / run	same as enrichment

¹ "Send data to table" is free when just moving data between tables. If the receiving table triggers new marketplace enrichment on arrival, those operations count separately at their own per-row Action and Data Credit cost.

Advanced: syncing a CRM out? Use Audiences — it's incremental (only changed records), versus a standard table import that re-pushes everything.

The part a PDF can't do

You now know the five leaks. Finding and fixing them in a live workspace is the job.

Doing all five inside a workspace that's actively running campaigns — without breaking the pipeline you're running on — is a different thing than reading about them. On a 30-minute screen-share I'll show you exactly where your workspace is bleeding. If we move forward, here's what I actually do:

Hunt down every silent re-run — which of your tables are re-enriching on a schedule nobody set, plus guardrails so one teammate's formula edit can't burn 60k credits overnight.
Right-size the model on every AI column — connect your LLM keys and match the model to the job: nano where it's enough, heavier where the research actually needs it. Cheaper and not dumber.
Connect and assign your data-provider keys per use case — map the subscriptions you already pay for to the right enrichment, so you stop buying data twice.
Build the filter-and-reuse layer — score down to your in-ICP rows before you spend a credit, plus a master-table setup so each company is enriched once and reused everywhere.
Rebuild your enrichment waterfalls provider-by-provider — cheapest source that still hits, per data type, so you stop paying 25 credits for a number you could've gotten for 2.

Plus the parts this playbook leaves out: whether you're on the wrong plan tier for your real usage, and rebuilding CRM sync to push only what changed.

The call is free, and so is the diagnosis — I'll show you the top leaks live whether or not you hire me. If your workspace doesn't have enough waste to pay for the fix, I'll say so on the call.

Get the 5 Ways done for you →

or reach me at victor@silvergtm.com.