Salesforce Headless 360 pricing in 2026. How to budget before the price list exists.

Salesforce has published no standalone Headless 360 production price. The cost runs through Agentforce Flex Credits and per user add ons, so you budget from those meters and negotiate the four lines still left blank before you sign.

How much does Salesforce Headless 360 cost in 2026?

Salesforce has published no standalone production price for Headless 360, and we do not expect one. Headless 360 opens the platform to agents through APIs, MCP tools and CLI commands, so its cost runs through meters you already buy: Agentforce Flex Credits and per user add ons.

What you pay is your modeled agent traffic multiplied by the rates written into your order form. Salesforce announced Headless 360 on April 15, 2026, and as of September 2026 the published rates in the table below are still the only fixed inputs to that calculation.

What is already included, and what already bills?

Salesforce's Headless 360 architecture guide says basic MCP access comes with Enterprise Edition and above, with no extra Agentforce SKU. Access and usage are priced separately, though. Agentforce actions invoked through MCP and Data 360 queries through MCP draw Flex Credits today, and the user or service account behind each call must hold a valid Salesforce license.

Which Headless 360 pricing lines are still undisclosed?

Four lines remain open: the production runtime rate for agent traffic, the MCP tool call unit, the completion pass through for model tokens, and the rate lock term. Each is set deal by deal, and a first draft that leaves one blank asks you to sign before that number exists.

The production runtime rate

This is the price per credit that applies once agents run in production. When the order form leaves it as a placeholder, it resolves later at Pay as you go, on the meter's terms, after your build has committed you to the traffic.

The MCP tool call unit

When an external agent reads or updates records or starts a flow through MCP, Salesforce plans to meter it under the two TBA usage types. Until the multiplier appears, the rate card cannot price that traffic, so settle the unit in the order form before outside agents use the tool surface at volume.

Our MCP tool call analysis prices that surface call by call.

The completion pass through

This is the cost of the model tokens consumed when agents generate responses, passed through to you. It grows with reasoning model adoption, because those models spend more tokens on each answer.

The rate card already lists four prompt tiers, Starter, Basic, Standard and Advanced, which Agentforce 1 Edition customers and Agentforce for Sales, Service and Industries customers consume in addition to actions. Ask which tier your agents' model calls fall into, what happens when you bring your own model, and what ceiling applies.

The rate lock term

A named rate helps little if it holds only for year one. Ask for the rate to hold for the full subscription term, with a cap on the increase at renewal, so the price you model is the price finance pays in year three.

Why we advise against waiting for a published price list

A common suggestion is to hold off on Headless 360 commitments until Salesforce publishes production rates. We disagree. The rates are set per deal, and the buyers who fill the blanks on their own terms set the market later buyers pay. By the time a list appears, your developers will have built against the platform.

Pilot now instead, in sandboxes and at small volumes, while you negotiate named rates for production. You keep the option to walk away, and the pilot gives you real volumes to argue from.

How do you budget for Headless 360 before prices are published?

Build the budget bottom up from each workflow, convert it at the published credit rate, and stress test it at peak volumes rather than averages. The model that held up across our reviews has five inputs:

1. Actions per task. Count the actions each agent design takes to finish one task, at 20 credits each.
2. Tasks per day. Use the volume at target adoption, since pilot volumes understate it.
3. Voice share. Price voice actions separately at their 30 credit rate. Our Voice licensing guide covers how Salesforce licenses voice beyond the action rate.
4. Agent to agent traffic. Add the calls agents make to other agents and tools. The headless architecture exists to produce this traffic, so expect it to grow fastest.
5. Supervising seats. Add the $125 add on for the people who review, approve or correct agent work.

The Headless 360 licensing pillar explains the meter mechanics behind each input.

A worked example for one service operation

Say a service organization runs a case triage agent that takes 5 standard actions per case, across 3,000 cases per business day and 21 business days a month. A voice channel adds 500 calls a day at 4 actions each, and a third of cases hand off to a second agent for 3 more actions.

At average volume that is 105,840,000 credits a year, or $529,200, plus $90,000 for the seats. Size any Pre-Commit pool near that annual average, because unused credits expire at term end, and budget cash for peak months like the $66,150 one.

The table also leaves out the two TBA usage types. Every record the agents read or update and every flow they start will add credits once Salesforce sets a multiplier, so keep those rows in your model at zero and fill them in the day the rate card changes.

Why should the model run high?

Account team estimates have run low. In our reviews, the credit forecasts account teams supplied came in 40 to 70 percent below realized consumption once agents went live. The forecasts were built on chat volume, while the non interactive traffic between agents and systems compounded underneath.

Applied to the example, a forecast 40 percent low would have shown $317,520 a year against $529,200 of real consumption, and one 70 percent low would have shown $158,760. Against that record, a peak volume model is the realistic one.

Who needs the $125 add on?

Buy the add on for the supervising cohort, the people who touch agent workflows, never for your whole license population. Salesforce sells it at $125 for Sales, Service and Field Service and at $150 for Industries Clouds, both with unmetered Agentforce use for employees.

Staff who use agents only now and then can sit on the $5 per month Agentforce User License, which meters their use against Flex Credits. In the example, moving 20 of the 60 supervisors to that license would save $2,400 a month before their metered use.

How can you check what your agents actually consume?

• Digital Wallet. It shows Flex Credit consumption in near real time and supports threshold alerts. Set an alert at your modeled monthly figure for every pilot.
• Sandbox rates. The rate card charges 16 credits per standard action and 24 per voice action in sandboxes, against 20 and 30 in production. Test cycles cost money, so put them in the budget.
• Your own record and flow counts. Salesforce does not meter record operations or process invocations yet, so count them yourself during the pilot from your agents' call logs, or from Event Monitoring API logs if you license it. You can then price that history the day a multiplier is published.
• The consumption model on your org. Salesforce does not support Flex Credits and $2 per conversation pricing in the same org. Confirm which one you are on before you model anything.

Why is a capped runtime rate the strongest term in a Headless 360 deal?

In the deals we reviewed, one clause separated the outcomes: a named production runtime rate, capped and locked for the term, against a placeholder that resolved to Pay as you go. Buyers who locked the rate cut their effective cost per action by 20 to 35 percent.

A rate agreed before the build starts is negotiated against alternatives, including not building at all. A rate accepted after the agents ship is negotiated against sunk cost.

Salesforce sells Flex Credits on two buying models, PayGo and Pre-Commit. Unused credits do not roll over into the next subscription term, and overages bill monthly in arrears at the contracted rate, so the contracted rate matters more than the size of the pool.

Which contract terms should you ask for?

• A named rate per credit. Write it into the order form for production agent traffic, so no placeholder can resolve to list price later.
• The same rate for overage. Peak months should bill at the contracted rate with a ceiling, never at Pay as you go.
• A lock for the full term. The rate holds every year of the subscription, with a cap on the increase at renewal.
• A ceiling on the TBA usage types. Agree the multiplier for record operations and process invocations now, or a maximum that applies when Salesforce publishes one, and extend it to any usage type that replaces them. That way a 30 day notice cannot reset your cost.
• The completion pass through in writing. Who pays for model tokens, at which rate, and up to what ceiling.
• Carryover inside the term. Unused credits from year one should carry into year two, and the order form should state what happens to any balance at term end.
• A definition of metered traffic. List which API and MCP calls count as agent traffic, so existing integrations stay outside the meter.

The Headless 360 versus Agentforce boundary decides which meters your traffic draws, and the Agentforce pricing guide covers the buying models underneath. Our list of Salesforce contract clauses shows how these terms sit alongside the rest of the order form.

What did our order form reviews show in 2025 to 2026?

Across the Agentforce and Data Cloud order forms we reviewed in 2025 and 2026, the gaps were structural. The same lines were missing or vague from one deal to the next.

• Seats priced, meter left open. Most first drafts priced the add on seats clearly and left the production consumption meter vague or empty. That is backwards, because the consumption meter is where the money is.
• The token line went unexamined. The completion pass through was the least understood item on every order form where it appeared, and the least negotiated.
• Named rates beat placeholders. Buyers who insisted on a named, capped rate before signing paid materially less per action than those who accepted the blank.

Of the four gaps, the unwritten token terms deserve the closest attention, because that line grows with reasoning model adoption as well as with task volume.

What will the Salesforce account team say, and how should you reply?

Expect the same few lines in most Headless 360 conversations. Each has a precise answer.

• "Pricing isn't final, so we'll leave the runtime rate open for now." Reply that you will sign with a ceiling rate in the order form and a clause that lowers it if Salesforce later publishes something cheaper.
• "MCP access is included in Enterprise Edition." Reply that access is included, yet Agentforce actions and Data 360 queries through MCP already draw credits and the TBA usage types will add more, so those need a price now.
• "A bigger Pre-Commit earns a better rate per credit." Reply that unused credits do not roll into the next term, so you will commit near your modeled average and take a capped rate for overage.
• "Give everyone the add on so usage is unmetered." Reply that you will license the supervising cohort and keep occasional users on metered access.
• "You'll get 30 days' notice before metering starts." Reply that 30 days is too short to redesign agents already in production, so the multiplier, or a ceiling on it, belongs in the contract now.

What to do next

1. Now. List every workflow you plan to expose through Headless 360 and count actions per task, voice share and agent handoffs.
2. During the pilot. Run it in a sandbox with Digital Wallet alerts on, and log your own counts of record operations and flow or Apex calls by external agents.
3. Before the first quote. Finish your peak volume model so your own credit figure exists before the account team sends theirs.
4. When the first draft arrives. Mark every open line, from the runtime rate and MCP unit to the token terms and rate lock, and send back named, capped numbers.
5. Before signature. Cut the add on quantity to the supervising cohort and confirm carryover and overage terms.
6. At any stage. The Salesforce practice can work through the order form with you, blank by blank.

Frequently asked questions