Headless 360, pricing the agent gateway into Salesforce
Headless 360 lets agents and code consume Data 360, Customer 360, Agentforce, and Slack through APIs and more than 60 MCP tools instead of the browser. Salesforce announced it in April 2026 without a production price list, and the absence is the point: the cost flows through meters you either pin before the build or discover after it.
Prepared by Redress Compliance · August 6, 2026 · Salesforce negotiation advisory. Based on 30 to 40 Agentforce and Data Cloud negotiations run 2025 to 2026.
Executive summary
Announced at TrailblazerDX on April 15, 2026, Headless 360 is the architectural turn: the platform consumed by agents and code through APIs and 60 plus MCP tools rather than by humans through screens.
Salesforce has published no standalone production price list, and none is coming, because Headless 360 is not a SKU. The cost flows through Agentforce Flex Credits and per user add ons that already exist.
The meters are concrete even where the packaging is not: an Agentforce action draws 20 Flex Credits, a Voice action 30, at roughly $500 per 100,000 credits, which prices a standard action near 10 cents and a Voice action near 15.
The per user add on runs about $125 per user per month, $150 in regulated industries. Three buying models wrap the credits: Pre purchase, Pay as you go, and Pre Commit, and the spread between them is where the money moves.
The forecasting record is one sided. Across our negotiations, account team credit burn forecasts ran 40 to 70 percent below actual consumption once agents went live, anchored on interactive chat volume while agent to agent and Voice traffic doubled the draw within two quarters.
The gap was widest exactly where teams believed the AI capability was simply included.
The most contested line is also the most often blank: the production runtime rate for agent traffic, left unpriced in first order forms.
Buyers who modeled their own traffic in credits and locked a capped unit rate for the term paid 20 to 35 percent less per action than buyers who signed on Pay as you go and negotiated after the meter was running.
What Headless 360 actually is, and is not
Headless 360 exposes the platform's four surfaces, Data 360, Customer 360, Agentforce, and Slack, to external agents and code through APIs and MCP tools. It is a consumption architecture, not a product purchase: your agents, or anyone's agents, calling Salesforce capabilities as tools.
The Headless 360 versus Agentforce comparison draws the boundary with the in platform agent product, and the MCP tool call cost analysis prices the tool surface call by call.
No price list is not no price. Every Headless 360 call lands on an existing meter: Flex Credits for agent actions, API entitlements, Data 360 consumption, and the per user add ons.
Announcing the architecture without a rate card moves the pricing conversation into your order form, which is either a risk or an opportunity depending on who arrives prepared.
The meters, credit by credit
| Meter | The rate | The exposure pattern |
|---|---|---|
| Agentforce actions | 20 Flex Credits per action, near 10 cents at the $500 per 100,000 rate | Agent to agent chains multiply actions per task far beyond chat volume |
| Voice actions | 30 Flex Credits, near 15 cents | Voice traffic doubled estates' draw within two quarters of launch in our file |
| Per user add on | About $125 per user per month, $150 regulated | Sized to the full population when only a cohort touches agent workflows |
| The runtime rate | Often blank in the first order form | The unpriced line that Pay as you go fills in later, at list |
The three buying models allocate risk: Pre purchase buys a credit block up front, Pay as you go floats at rack rates, and Pre Commit trades a committed volume for a unit rate.
The record is unambiguous: modeled traffic plus a Pre Commit with a capped unit rate beat Pay as you go by 20 to 35 percent per action, and the cap mattered more than the commit, because the cap is what survives the growth.
The wider credit mechanics sit in the Agentforce pricing guide and the Agentforce pillar.
The Agentforce Flex Credits playbook
The credit forecasting worksheet, the buying model comparison, the capped rate clause language, and the consumption governance for agent estates.
Get the white paper →Forecasting burn, why the account team's number is half yours
The 40 to 70 percent forecast gap in our file was not sales malice; it was anchoring.
Forecasts started from interactive chat volume, the traffic humans generate, while the architecture exists precisely to generate the other kind: agent to agent calls, retries, tool chains, and Voice, none of which correlate with chat history.
Two quarters after launch, the non interactive traffic had doubled the draw in the estates we tracked.
The buyer side forecast is built from the workflow, not the history: actions per task per agent design, tasks per day at target adoption, the Voice share priced at its premium, and a growth curve for the agent to agent traffic the roadmap implies.
Model it in credits, convert at the negotiated rate, and the number you carry into the order form is the one the meter will eventually confirm.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across Agentforce negotiations, 2025 to 2026
Across roughly 30 to 40 Salesforce Agentforce and Data Cloud negotiations Morten Andersen and the team ran in 2025 and 2026, the consumption layer followed the same script:
Account team credit forecasts against realized consumption once agents went live, anchored on chat volume while agent traffic compounded.
Per action saving for buyers who locked a capped unit rate for the term versus those who signed Pay as you go.
The structural pattern was sequencing: consumption meters introduced late in the cycle, after the seat count was agreed, when the negotiating leverage was spent.
The estates that did best inverted it, pricing the consumption layer first, because seats are predictable and meters are not, and the meter is where the next three years of surprise lives.
Your first five moves
- Model agent traffic in credits before the build: actions per task, tasks per day at target adoption, the Voice share at its premium rate, and the agent to agent growth curve.
- Fill in the blank line. The production runtime rate goes in the order form, capped, for the term, before anything ships.
- Choose Pre Commit with a cap over Pay as you go wherever modeled traffic supports it; the cap outlives the commit.
- Scope the per user add on to the cohort that touches agent workflows, not the license population.
- Negotiate the consumption layer before the seat count, while the leverage still exists. The Salesforce practice runs the sequence with you, and the pricing expectations analysis tracks what Salesforce publishes next.
Frequently asked questions
What is Salesforce Headless 360?
The architecture Salesforce announced at TrailblazerDX on April 15, 2026: agents and code consuming Data 360, Customer 360, Agentforce, and Slack through APIs and more than 60 MCP tools rather than through the browser.
It is a consumption pattern, not a SKU, and its cost flows through existing meters, chiefly Agentforce Flex Credits.
How much does Headless 360 cost?
There is no standalone price list. Traffic draws Flex Credits, 20 per Agentforce action and 30 per Voice action, at roughly $500 per 100,000 credits, near 10 and 15 cents per action, plus a per user add on around $125 per month, $150 in regulated industries.
The realized cost is your traffic times the unit rate you negotiate.
What are the Agentforce buying models?
Pre purchase, a credit block bought up front, Pay as you go at rack rates, and Pre Commit, committed volume for a negotiated unit rate.
In our negotiations, modeled traffic plus a Pre Commit with a capped unit rate beat Pay as you go by 20 to 35 percent per action, and the cap mattered more than the commit size.
Why do Agentforce credit forecasts run so low?
They anchor on interactive chat volume, while the architecture generates non interactive traffic, agent to agent calls, retries, tool chains, and Voice, that compounds independently.
Account team forecasts ran 40 to 70 percent below realized burn in our file, with Voice and agent to agent traffic doubling the draw within two quarters of launch.
Is AI usage included in Salesforce licenses?
The capability ships in the platform; the consumption meters. The belief that agent usage was simply included produced the largest forecast gaps in our negotiations, because the credits drew silently against Pay as you go rates.
Every agent workflow should be modeled in credits and priced in the order form before it ships.
When should the Headless 360 pricing conversation happen?
Before the build, and before the seat negotiation. Consumption meters introduced late in the cycle, after seats were agreed, met spent leverage in every negotiation we ran. Price the runtime rate, the cap, and the buying model first; the seat count is the predictable half of the deal.
