Do Sandbox Agentforce Actions Burn Flex Credits, and at What Rate Versus Production?

The buyer question, stated cleanly: do sandbox and other pre-production Agentforce actions burn Flex Credits, and what is the sandbox multiplier versus production? RevOps, GTM engineering, enablement, and revenue finance keep asking it because pilot budgets are approved on production math while the actual burn starts in UAT.

The answer is yes, and it is published. Salesforce's Flex Credits Rate Card, updated 08.31.2026, applies a sandbox multiplier to pre-production environments, including sandboxes and scratch orgs. Source: Salesforce Flex Credits Rate Card 08.31.2026.

This page is a decoder, not a price list. Order form beats any blog math. Every figure below is labeled with its source, and nothing here is a quote.

What the Rate Card actually says

Two lines matter for planning. Source for both: Salesforce Flex Credits Rate Card 08.31.2026.

• Standard and custom Agentforce actions: 20 credits in production, 16 credits in sandbox.
• Standard and custom voice actions: 30 credits in production, 24 credits in sandbox.
• The sandbox multiplier applies to pre-production environments generally, which the card describes as including sandboxes and scratch orgs. A scratch org is not a loophole.

Two secondary explainers read the card the same way. Aquiva describes sandbox usage as billed at 80% of the production rate, which it puts at 16 and 24 credits. Source: Aquiva. AgencyQ states 16 credits in sandbox against 20 in production and recommends debugging token overflows in sandbox rather than production. Source: AgencyQ.

Treat the 80% framing as a convenient way to remember the two published pairs, not as a rule you can apply to every future usage type. Re-read the current rate card before each planning cycle. Salesforce dates the card, and the date moves.

Why this opens a Proof Gap

The Proof Gap is the distance between what a team claims an AI deployment cost or produced and what it can evidence at close. Sandbox burn widens it in three predictable ways.

1. Teams treat sandbox as free. Build, test, and demo cycles run for weeks before anything is production-facing, and the credits are already gone.
2. The invoice mixes environments. Digital Wallet reporting is where consumption becomes visible, and if nobody separates environments the pilot appears more expensive than the production use case actually is.
3. UAT burn eats pilot budget. A long UAT on a multi-action agent at 16 credits a call can consume the credits reserved for the production proof, so the pilot ends without the numbers finance asked for.

The failure is not the sandbox rate. The failure is a pilot budget approved on production arithmetic that never had a pre-production line in it.

The RevOps operating rule

1. Open a separate Digital Wallet line for sandbox. CloudAnalysts recommends monitoring sandbox and production consumption separately rather than reading one org-level number. Source: CloudAnalysts.
2. Set a UAT credit budget before the build starts, expressed in credits, with a named owner who can pause the test cycle.
3. Do token-overflow debugging in sandbox. AgencyQ makes the point directly: overflow retries are the expensive failure mode, and they are cheaper to reproduce in pre-production.
4. Price UAT at 16 credits per standard action and 24 per voice action for planning, then replace both with your contracted rate as soon as the order form exists.
5. Reconcile sandbox against production at close. If the two lines are not separable in your month-end pack, the pilot cost claim is unverifiable.
6. State the sandbox line in the board slide. A pilot cost that omits pre-production is not a pilot cost.

What this page is not

• It is not the arrears forecast page. That model lives at Flex Credits forecast and close.
• It is not the kill-switch page. Digital Wallet alerts without blocking, covered at Digital Wallet has no kill switch.
• It is not the chargeback page. Org-level allocation is covered at BU chargeback.
• It is not the buying-model fork. Flex Credits versus the $2 Conversation SKU is at the break-even worksheet.
• It is not a price list. Credit counts are rate-card units, not dollars, and your contracted rate governs.

Related: Flex Credits forecast and close · Digital Wallet · BU chargeback · Break-even worksheet · AI pricing and contracts · Agentforce and Data 360 share one credit pool