TL;DR
- Clay separates Actions, which measure workflow execution, from Data Credits, which pay for marketplace data and AI. An external API key can reduce the second bill, but an enrichment still uses an Action.
- Clay lists Launch from $185 per month on monthly billing and Growth from $495 per month on monthly billing, with lower annual rates. Launch includes 15,000 Actions and Growth 40,000; confirm the entry Data Credit allowance because Clay’s pricing page presents conflicting figures.
- A workload of 1,000 records, three enrichments and one CRM export per record consumes 4,000 Actions if every operation runs. At 10,000 records it consumes 40,000, before retries or other work. A CSV download is different: Clay says it does not consume Actions.
- The subscription is only one part of the decision. Add marketplace credits, outside provider bills, unused allowances and the work required to maintain the flow. Choose based on cost per accepted record, not price per credit in isolation.
- Launch is the first tier to investigate for a straightforward small workflow. Growth becomes relevant when the action allowance or the required integrations demand it. A cheaper data API may fit better when the application already owns the workflow and Clay's operating environment adds little value.
What does a Clay subscription pay for?
Clay gives a team a place to find, enrich, research and activate accounts and contacts. The commercial decision is whether that shared operating environment saves enough work to justify its subscription and usage costs. A sales operator managing a visible workflow and an engineer embedding enrichment in a product may come to different answers even with identical data requirements.
Clay's pricing page lists the following starting allocations. They are entry points, not a quote for every combination of usage, billing term and features.
| Plan | Starting price | Included Actions per month | Included Data Credits | Buying question |
|---|---|---|---|---|
| Free | $0 | 500 | 100 | Does a small pilot justify upgrading? |
| Launch | $185/month, billed monthly | 15,000 | Confirm allowance: 2,500 and 3,000 appear in different places | Can the workflow fit its capacity and feature set? |
| Growth | $495/month, billed monthly | 40,000 | 6,000 | Do we need the extra capacity or Growth integrations? |
| Enterprise | Contact Clay | Confirm in quote | Contract dependent | What capacity, controls and terms are in the proposal? |
Clay also publishes annual-billing equivalents of $167 per month for Launch and $446 per month for Growth, subject to the annual commitment. On the monthly-billing basis used here, the $310 difference between their starting prices buys more than an increment of data. Growth also lists capabilities such as CRM auto-sync, HTTP API integrations and webhook automation. If the workflow depends on one of them, compare the applicable plan and implementation design before choosing on volume alone. Verify current plan terms and whether the advertised starting price assumes a particular billing arrangement.
Our Treg vs Clay comparison tackles the adjacent architecture choice: where the workflow and its review process should live. Here the question is how to budget work that runs in Clay.
Actions and Data Credits are different budgets
Clay's usage documentation describes an Action as platform work, such as an enrichment, AI run or record sent to an external system. A Data Credit pays for eligible data or AI purchased through the marketplace. The exact data-credit charge varies by operation and provider.
| Operation | Action treatment | Data Credit treatment |
|---|---|---|
| Import existing first-party records | No Action for the import | No Data Credits for the import |
| Enrich a record through a marketplace provider | One Action per enrichment | Varies with the selected data and provider |
| Enrich with your own provider key | One Action per enrichment | No Clay Data Credits for that external provider; its own bill still applies |
| Run an AI operation | An Action for the run | Marketplace credits or a separate provider charge when using your own key |
| Export a record to a CRM or external sequencer | One Action per record | No Data Credits for that export |
| Download a CSV | No Action | No Data Credits |
The distinction between a CRM export and a CSV download matters. “Export” is too broad a line item for a budget: one path consumes platform capacity and the other does not. The table describes Clay's documented units; check the actual operation in your workspace before committing a production budget.
Actions do not roll over. Clay documents monthly-plan Data Credit rollover up to twice the monthly allowance, with different terms for annual plans and renewals. A buyer comparing an annual offer should ask how its own unused balance expires rather than applying the monthly rule automatically.
What does a 1,000-record workflow cost?
Take a constructed example: a revenue team imports 1,000 company records, runs three separate enrichments on every record, then sends every record to its CRM. This is an accounting exercise, not a test of provider coverage.
| Step | Runs | Actions per run | Actions |
|---|---|---|---|
| Import existing records | 1,000 | 0 | 0 |
| Enrichment A | 1,000 | 1 | 1,000 |
| Enrichment B | 1,000 | 1 | 1,000 |
| Enrichment C | 1,000 | 1 | 1,000 |
| CRM export | 1,000 | 1 | 1,000 |
| Total | 4,000 |
That uses about 27% of Launch's stated 15,000 monthly Actions, leaving 11,000 before other workspace activity. It does not prove the workflow fits the included Data Credits. Clay’s pricing page shows 2,500 in its plan summary but 3,000 in a selector; confirm the allowance for the exact configuration before purchasing. To estimate consumption, list the credit charge for each selected enrichment and multiply it by the number of chargeable results under that provider's rules. Phone numbers, AI research and fallback providers can change the mix substantially.
If only 600 records pass the team's acceptance rule and reach the CRM, the export contributes 600 Actions rather than 1,000. Under the same three-enrichments-for-every-record assumption, the total becomes 3,600 Actions. Whether failed or empty enrichment results consume credits or receive a refund depends on the operation and provider; verify those terms rather than treating all 3,000 attempts as successful purchases.
Now divide the complete monthly cost by the 600 accepted records. If the required export method is available on Launch, a $185 Launch subscription alone represents about $0.31 per accepted record at that volume, before extra marketplace credits, outside provider charges and staff time. The action arithmetic does not establish that Launch supports the required CRM sync; if Growth is needed, its $495 starting subscription alone is $0.825 per accepted record under the same allocation. If the same subscription also supports other useful workflows, allocate it consistently instead of assigning its entire cost to this one job by default.
What changes at 10,000 records?
At ten times the volume, the same unconditional workflow requires 30,000 enrichment Actions plus 10,000 CRM-export Actions: 40,000 Actions. That matches Growth's published starting allowance before any reruns, additional research, signals or other teams' work. It is a capacity boundary, not a comfortable operating target.
A conditional flow can be materially different. Suppose all 10,000 records receive the first enrichment, only 6,000 need the second and 3,000 need the third, while 6,000 accepted records go to the CRM. The arithmetic becomes 10,000 + 6,000 + 3,000 + 6,000 = 25,000 Actions. Those counts are assumptions, not Clay performance measurements. The point is that the shape of the workflow matters as much as the input list size.
At the $495 Growth starting price, allocating the whole base subscription to 6,000 accepted records gives roughly $0.083 per accepted record, before data and operating expenses. If the marketplace bill adds $300 in a hypothetical month, the combined $795 is about $0.13 per accepted record. The $300 is an illustrative input, not a published Clay price or a prediction of what this job costs.
Do not assume a 40,000-Action workload can run safely on exactly 40,000 Actions when the same workspace performs other jobs. Estimate a reserve from measured reruns and other activity, then confirm the tier and overage path with Clay. Its documentation says Data Credits can be topped up, whereas Actions are tied to capacity and require a tier adjustment.
Does bringing your own API key save money?
It can, particularly if your team already pays for a provider whose unused allowance covers the same lookups. Clay says using an external key avoids Clay marketplace Data Credits for that provider, while the Action remains. The provider may charge by request, successful match, returned field, monthly minimum or annual commitment. Include the actual contract and its limitations.
Consider the 1,000-record example. Assume the third enrichment would otherwise use one marketplace Data Credit per chargeable result. If all 1,000 results were chargeable and an external key replaced that purchase, the Clay ledger would show 1,000 fewer Data Credits. The workflow would still use 4,000 Actions under the original assumptions. Whether the change saves money depends on the provider's incremental invoice, any minimum commitment and the value of capacity already paid for.
Do not compare a marketplace credit with a provider's list price without checking what each returns. An email, verified email and mobile number are different outputs. A lower price on a partial record can be more expensive if the team must buy another lookup to make it usable.
A separate enrichment API can be a better fit when software already owns matching, retries, data storage and CRM updates. Clay can be the better purchase when operators need to inspect, change and run multi-provider workflows themselves. The trade-off is not simply “credits versus API calls”: it is platform subscription and operator control versus the engineering and maintenance of the alternative. See the waterfall enrichment guide for the workflow design questions behind that calculation.
Build an invoice you can reconcile
Use one row per operation and keep the units separate. This ledger works in a spreadsheet without assuming every attempted lookup returns a useful field.
| Column | Example entry | Why it belongs |
|---|---|---|
| Input and operation | 1,000 contacts; email lookup | Defines the job rather than the provider's marketing unit |
| Actual executions | 1,000 | Includes retries and excludes rows filtered before running |
| Actions | 1,000 | Reconciles platform capacity |
| Chargeable marketplace results | 700 | Depends on the provider's actual charging rules |
| Data Credits per chargeable result | 1, assumed | Replace with the rate shown for the chosen integration |
| External provider bill | $0 or actual invoice | Captures a BYOK subscription and usage outside Clay |
| Accepted records | 600, assumed | Denominator for the buyer's actual outcome |
For this illustrative row, 700 chargeable results at one Data Credit each would use 700 Data Credits. It says nothing about whether those results are accurate or usable. Record the result, its source and the acceptance rule separately. The same contact can consume several operations; it should not be counted several times as an accepted contact.
At the end of the billing period, reconcile the workspace's Actions and Data Credits with this ledger, then add the external provider invoice. Explain any difference from the original estimate: automatic reruns, additional fields, extra exports and duplicates are common places to investigate. Clay's usage page can help locate consumption, but external provider costs must be checked in that provider's account.
Which plan fits the work?
Start with Launch when the team needs an operator-facing workflow, the required features are available on that tier and a measured workload fits within its Action and data budgets. Do not upgrade solely because an annual capacity calculation looks close: check the busiest actual billing month and the work already sharing the workspace.
Evaluate Growth when the workflow needs its listed CRM sync, HTTP or webhook capabilities, or when recurring work requires more Actions than Launch provides. The meaningful comparison is the incremental subscription plus data purchases against the operational benefit, not the nominal cost of 25,000 additional Actions.
Request an Enterprise proposal when capacity, controls, security or procurement terms exceed the self-serve plans. Ask for the action and data allocations, the treatment of unused credits, any minimum term and how usage changes during the contract. A custom quote is a starting point for comparison, not proof that an enterprise plan is necessarily cheaper at scale.
Consider an external API-led build when the system already contains the orchestration, audit trail and human review Clay would otherwise provide. Include engineering ownership, failure recovery and staff time in the model. If those components have to be built from scratch, the apparent saving can disappear before the first useful record reaches sales.
FAQ
Does using my own API key make Clay free?
No. Clay says the external provider's usage is billed separately and the Clay enrichment still consumes an Action. The subscription and any other marketplace purchases remain part of the total.
Does downloading a CSV use an Action for each row?
Clay's documentation says a CSV download does not consume Actions or Data Credits. Sending records to a CRM or an external sequencer is a different operation and is documented as consuming an Action per record.
Can I buy more Data Credits without upgrading?
Clay documents one-time and automatic Data Credit top-ups for eligible plans, with a premium on top-ups. Actions are handled differently and cannot be purchased as one-time top-ups under the documented model. Check current eligibility, price and limits before relying on either path.
What happens to unused credits?
Actions reset without rolling over. Clay describes a capped rollover for unused monthly-plan Data Credits and separate renewal conditions for annual plans. Check your contract and current billing screen before treating a balance as permanent.
What should I measure in a pilot?
Use the same input set and acceptance criteria across the candidate workflows. Record operations run, Actions, marketplace credits, outside invoices, returned fields and accepted records. Test a small batch first, then use the results to estimate the peak month rather than claiming a sample's match rate will hold for every market.





