Agentforce Pricing: Agentforce Cost for Service and Sales, and How to Estimate Your Bill

Agentforce is sold six different ways at once, so there is no single Agentforce price. On the US pricing page, Flex Credits cost $500 per 100,000, a standard agent action burns 20 credits ($0.10) and a voice action burns 30 ($0.15). Conversations are $2 each. An Agentforce User License is $5 per user per month plus credits, a flat add-on is $125 per user per month, and Agentforce 1 Editions start at $550 per user per month. To estimate your bill you need one number Salesforce cannot give you: how many actions your average conversation triggers.

Re-verified against salesforce.com/agentforce/pricing on 4 September 2026. All rates below still stand. For the full line-by-line rate card, including the three buying models and the two FAQ rules that decide a credit purchase, see our breakdown of Agentforce pricing.

Last updated August 2026. Re-verified against Salesforce's US pricing page on 25 August 2026.

Every figure below was read directly from Salesforce's own Agentforce pricing page, first on 18 August 2026 and re-checked on 25 August 2026, with the currency selector set to US Dollar. Nothing had changed between those two readings. Salesforce prints a disclaimer on that page saying it is provided for information purposes only and is subject to change, so treat these as a starting point for a quote rather than a contract.

How much does Agentforce cost?

There is no one answer, because Salesforce publishes six buying models for the same product. Usage models price the work: Flex Credits at $500 per 100,000 credits, or Conversations at $2 each. Seat models price access: $5 per user per month for a user license that still needs credits, $125 per user per month for the unmetered add-on ($150 for Industries), and from $550 per user per month for Agentforce 1 Editions, which bundle the add-on with 2.5M Flex Credits per org per year. Salesforce Foundations lets you build and test at $0.

How much does Agentforce for Service cost?

Agentforce for Service is not priced as a separate product with its own rate card. It is the same Agentforce runtime pointed at customer service, so it draws on whichever of the six models you bought. In practice most service deployments land on one of two: Conversations at $2 per customer conversation, or Flex Credits at $500 per 100,000 credits where a standard action costs 20 credits, meaning $0.10 an action. A support conversation that fires five actions costs $0.50 on credits against $2.00 flat on Conversations.

The wrinkle specific to service is Voice. A Voice action costs 30 credits rather than 20, so $0.15 instead of $0.10. If you run a phone channel through Agentforce, model your credit burn at the higher rate or you will underestimate by half again on every voice interaction.

How much does Agentforce for Sales cost?

Agentforce for Sales pricing uses the same answer, the same six models and a different usage shape. Sales agents such as the SDR and Sales Coach agents consume Flex Credits per action or bill per Conversation exactly as service agents do. The difference is that sales work tends to be action-heavy rather than conversation-heavy: research a lead, enrich a record, draft an email, log the activity, book the meeting. Five or six actions on a single outbound touch is normal.

That pushes sales use cases toward the Conversations model faster than service use cases. At 20 credits an action, a six-action sales sequence costs $0.60 on credits, still under the $2.00 Conversation rate, but a genuinely autonomous SDR working a lead over several days crosses twenty actions quickly and the flat $2.00 becomes the cheaper unit.

How much does Agentforce cost per month?

Salesforce's own five worked examples run from $10 to $1,800 a month, which is a 180-fold spread on the same product. Those examples consume 2,000, 24,000, 36,000, 60,000 and 360,000 Flex Credits respectively. For a B2B SaaS company the honest planning range is narrower: a small support team handling a few thousand conversations a month sits in the hundreds, while a team running both service and sales agents at volume sits in the low thousands.

Two caveats before you take those numbers to a budget meeting. Salesforce footnotes that its worked examples exclude Data 360 credits, so treat every published figure as a floor rather than a total. And if you buy seats instead of usage, the arithmetic inverts completely: the $125 per user per month unmetered add-on costs a fifty-person team $6,250 a month regardless of how little the agents actually do.

What is a Flex Credit in Agentforce?

A Flex Credit is Salesforce's consumption unit. In its own words, each action is "a specific function that an AI agent executes on the platform, such as updating a record, summarizing a complex case, answering a product inquiry, or executing a custom prompt or flow" and every action draws from a shared credit pool. Standard actions cost 20 credits. Agentforce Voice actions cost 30.

Do the division and one credit is worth half a cent. That makes a standard action $0.10 and a voice action $0.15. Credits are pooled across teams, channels and use cases, which is genuinely useful: you are not forecasting per department, you are forecasting one number for the org.

Is Agentforce cheaper per action or per conversation?

The break-even is exactly twenty actions. Twenty standard actions cost 20 x 20 credits = 400 credits = $2.00, which is precisely the per-conversation price. So conversations averaging fewer than twenty actions are cheaper on Flex Credits, and long multi-step conversations are cheaper at the flat $2 conversation rate.

That single number is the whole estimating exercise, and it is the one thing you cannot look up. A password reset might be two actions. A case that reads an order, checks a shipment, updates a record, drafts a reply and summarizes the thread is five before anyone follows up. Salesforce's own worked examples on the pricing page use two, three and five actions per use case, which suggests most service work sits comfortably under the twenty-action break-even and therefore favors Flex Credits.

Can you use Flex Credits and Conversations at the same time?

No. Salesforce's own FAQ states plainly that Flex Credits and Conversations will not be supported in the same org. You pick one consumption model for the entire organization, and you live with it. You can move from Conversations across to Flex Credits later, but it means swapping everything over rather than running the two side by side, so this is a decision to get right at signing rather than one to revisit next year.

There is a second cost to that choice that is easy to miss. Salesforce lists granular agent analytics as available only with Flex Credits. Picking the simpler flat $2 conversation rate therefore trades away reporting depth, which is an odd thing to give up on the exact product whose value you will be asked to prove to a finance team. Flex Credit consumption is tracked in Digital Wallet, with near real-time usage and threshold alerts. Credits are also fungible across actions, prompts, translations and voice actions, so one pooled budget covers the whole org.

What do Salesforce's own worked examples cost?

The clearest thing on the pricing page is the set of worked calculations Salesforce publishes itself. They are worth reading before you build your own model, because they show the real spread. Reproduced with Salesforce's arithmetic intact at $500 per 100,000 credits:

Same product, same rate card, and a 180 times spread between the cheapest and the most expensive example. What moves the number is not the price per action, it is how many people trigger how many actions. Salesforce also prints a footnote saying these examples are illustrative and exclude Data 360 credits and other consumption services, so treat every row as a floor rather than a quote.

How do I estimate my Agentforce bill before signing?

Pull thirty real transcripts from last month and count, for each one, how many discrete things an agent would have had to do: every lookup, every record update, every drafted reply, every summary. Average them. Multiply that average by 20 credits, by your monthly conversation volume, by $0.005. That is your Flex Credits estimate. Compare it against volume times $2 for the conversation model.

Worked example. A support team handles 3,000 AI-eligible conversations a month, averaging four actions each. That is 4 x 20 x 3,000 = 240,000 credits, or $1,200 a month on Flex Credits. The same volume on the conversation model is 3,000 x $2 = $6,000. Flex Credits win by five to one at that action count, and the gap only closes if your conversations get much longer.

Two cautions on that math. It counts the AI meter only, not the Salesforce seats underneath it, and most orgs buying Agentforce are already paying for Sales Cloud or Service Cloud licenses. It also assumes your action count is stable, and action counts tend to rise as teams give the agent more to do, which is the normal and healthy direction of travel.

What happens if I go over my Agentforce entitlement?

Salesforce states there is no overage penalty. Its exact wording: "If you exceed your entitlement, your rate is your contracted rate billed monthly in arrears." It points customers at Digital Wallet alerts to catch the trend early. That is more buyer-friendly than the punitive overage rates some AI vendors publish, but it still means the invoice moves with usage rather than sitting still.

One restriction worth knowing before you plan around a buying model: Salesforce says conversation-based pricing, the Agentforce add-ons and Agentforce 1 Editions are not available with the PayGo and Pre-Commit buying models. Flex Credits are. If you wanted to start pay-as-you-go and switch to per-conversation later, confirm the path with your account executive rather than assuming it.

How does Agentforce pricing compare to other AI support agents?

Every serious vendor now meters AI separately from seats, but they meter different things, which is why list prices are not comparable. Fin, the company formerly called Intercom, charges $0.99 per resolution and $9.99 per qualification, and does not charge when a conversation is simply passed to a human without an outcome. Help Scout charges $0.75 per AI resolution. Zendesk bills on outcomes but publishes no dollar rate at all. We broke the definitions down in our guide to what actually counts as a billable AI resolution, because the definition moves the bill more than the headline rate does.

The comparison that matters is not rate against rate. It is unit against unit. An action, a conversation, a resolution and a qualification are four different things, and a vendor charging a higher rate on a rarer unit can easily be cheaper. Our customer service software buyer guide prices an identical workload across six billing units for exactly this reason.

Should I buy Agentforce or a standalone AI agent?

Buy Agentforce if the work happens inside Salesforce. If your agent needs to execute flows, update arbitrary objects, respect your permission model and act on CRM records, it is the product built for that and nothing external matches it. Budget for an implementation project, because defining, permissioning and testing actions against your org's configuration takes weeks, not days.

Look at a standalone agent if the work happens where your customers message. Text, web chat, email, Instagram DM, Facebook Messenger and WhatsApp are not Salesforce surfaces, and an agent that lives on those channels and syncs the contact, the qualification and the booking back into the CRM is a different shape of tool. That is the case we make on our Salesforce chatbot integration and pricing page, alongside the honest limits: we do not run inside Service Cloud and we do not execute Apex.

There is a data question underneath both options that buyers routinely skip. An AI agent is only as good as the records it reads, and if the answer your customer needs lives in a warehouse, a billing system or an ops database rather than in the CRM, you will need a way to connect those systems to the tools that query them before any agent can answer accurately. Agent quality problems are frequently data plumbing problems wearing a costume.

Does the Salesforce acquisition of Intercom change the math?

Not yet, and not for your current contract. Salesforce signed a definitive agreement on 15 June 2026 to acquire Intercom, which had renamed itself Fin, for roughly $3.6 billion, with closing expected in the fourth quarter of Salesforce fiscal 2027. Existing Fin terms are unaffected by a signature.

What it changes is the shortlist. The two AI support agents US buyers most often evaluate against each other will sit under one owner, which narrows the field for anyone treating vendor independence as a procurement requirement. If that matters to your organization, raise it during this evaluation rather than at renewal. We track the detail in our breakdown of Intercom pricing and what Fin costs.

Three rules Salesforce puts in the FAQ that change the credit math

Re-reading the pricing page on 4 September 2026 turned up three constraints that sit below the rate card and matter more than any headline number.

Unused Flex Credits do not roll over into subsequent subscription terms. Whatever you pre-purchase and fail to consume expires at renewal. Given that Salesforce's own examples burn credits at rates differing 180 to 1, a twelve month pre-purchase is a bet on a forecast you cannot yet measure.

There is no overage penalty. If you exceed your entitlement, Salesforce bills at your contracted rate, monthly in arrears. Put that next to the expiry rule and the guidance is unambiguous: when your forecast is a range, buy near the bottom of it. Under-buying costs you only the volume discount. Over-buying costs you the whole unused balance.

The flexible buying models apply to Flex Credits only. Salesforce publishes three: Pre-Purchase, Pre-Commit and PayGo. Its FAQ then states that Flex Credits are available with PayGo and Pre-Commit, while Conversation-based pricing, the Agentforce add-ons and Agentforce 1 Editions are not. Pre-Commit also carries the note that it is widely available later this year, so on 4 September 2026 the live choices are Pre-Purchase upfront or PayGo in arrears. If a no-commitment start matters, that requirement picks your metering model before price does.

We work through the arithmetic on each of these, including what a single 100,000 credit pack buys against each of Salesforce's five published examples, in our guide to Agentforce Flex Credits pricing.

The short version

Agentforce is competitively priced for its category once you know your action count, and Salesforce deserves credit for publishing real numbers when several rivals publish none. The risk is not the rate. It is that five simultaneous buying models make the bill genuinely hard to forecast, and the one input that decides everything, actions per conversation, is something you only measure accurately after go-live.

If forecastability matters more to you than platform depth, the counter-position is a flat subscription with no per-conversation meter at all. Ours runs $79, $199 and $499 a month across every messaging channel, with carrier and Meta fees passed through at cost. If platform depth matters more, buy Agentforce, and go in with thirty transcripts counted so the quote is a negotiation rather than a guess. If you are still building a shortlist, we keep a running comparison of Agentforce alternatives. Either way, do the counting first.