Salesforce / Headless 360
Headless 360 vs Agentforce. How each bills.
Headless 360 is the infrastructure layer and Agentforce is the agent runtime. They bill on different meters. See where double counting hides and how to avoid paying twice.
Headless 360 is the infrastructure layer and Agentforce is the agent runtime. They bill on different meters. See where double counting hides and how to avoid paying twice.
Headless 360 is the infrastructure layer that exposes Salesforce to agents, and Agentforce is the runtime that executes them. They bill on different meters, and a headless design usually runs both. This guide shows how each bills, where double counting hides, and how to avoid paying twice.
Buyers often treat Headless 360 and Agentforce as one thing. They are not. One is the plumbing and one is the worker, and each carries its own meter.
Headless 360 is the infrastructure layer and Agentforce is the agent runtime, so they bill on different meters entirely. Salesforce describes the Agentforce runtime as the layer that executes actions.
Headless 360 exposes the platform to agents and code through APIs and more than 60 MCP tools. It is the access layer, and its cost tracks the data an agent reads and writes.
Agentforce runs the agents. It bills in Flex Credits per action, per the Agentforce pricing page. Its cost tracks activity, not data volume.
The infrastructure layer bills on data and access while the agent layer bills on actions, and the two move independently. A quiet agent that reads a lot of data can cost more on infrastructure than on credits.
Data 360 meters ingestion, storage, and processing. Every lookup and segmentation an agent triggers runs here, so infrastructure cost rises with data touched.
Agentforce meters Flex Credits at 20 per action and 30 per Voice action. Reasoning steps also draw credits, so a chatty agent burns credits fast regardless of how much data it reads.
Infrastructure layer versus agent layer
| Dimension | Headless 360 infrastructure | Agentforce runtime |
|---|---|---|
| Meter | Data 360 and access | Flex Credits |
| Scales with | Data touched | Actions taken |
| Unit | Ingest, store, process | 20 credits per action |
| Invoice line | Platform and data | Consumption |
Double counting hides when a single agent task draws both a Flex Credit for the action and a Data 360 charge for the lookup that action triggers. The two charges sit on different invoice lines.
An agent that resolves a customer question takes an action, which costs 20 credits, and queries Data 360 to ground the answer, which costs data processing. One task, two meters.
Most forecasts count only the agent layer because Flex Credits are the visible headline. The infrastructure draw underneath is easy to miss until the first invoice arrives.
The common advice is that once you own Agentforce, the headless infrastructure is just included platform capability, so you only need to watch Flex Credits. We disagree. In the estates we benchmarked, the Data 360 draw behind agent actions often matched the Flex Credit spend, so watching credits alone understated the true cost by roughly half. The buyer side move is to model both layers as one workload, forecast the data processing alongside the action burn, and negotiate a capped rate on each meter. Counting one layer and ignoring the other is exactly how the invoice surprises you.
Source: Redress Compliance advisory engagement file, 2025 to 2026.
The seat count tells you almost nothing about an agent estate. The credit line and the data line together tell you the truth.
The levers all come down to counting both layers and capping both meters before you build. A single blended forecast beats two separate guesses.
Negotiate a capped rate on Flex Credits and a defined rate on Data 360 processing. Salesforce set the platform direction in its newsroom, but the rates are yours to lock. Read the Headless 360 licensing pillar for the full layer map.
No. Headless 360 is the infrastructure layer that exposes Salesforce to agents and code through APIs and MCP tools. Agentforce is the agent runtime that executes actions and bills in Flex Credits. A headless design usually uses both, and each has its own meter.
The infrastructure layer bills through platform entitlements, API access, and Data 360 consumption for the data an agent reads and writes. It is the plumbing, and its cost rises with the volume of data an agent touches, not with the number of agent actions.
The agent layer bills in Agentforce Flex Credits, at 20 credits per action and 30 per Voice action, roughly 500 dollars per 100,000 credits. Every reasoning step and action an agent takes draws credits, so the agent layer scales with activity.
Double counting hides when a single agent task draws both a Flex Credit for the action and a Data 360 charge for the lookup it triggers. Teams that forecast only the agent layer miss the infrastructure draw underneath it and understate the bill.
Map each workload to both layers, forecast the Data 360 draw alongside the Flex Credit burn, and negotiate a capped rate on each meter. The goal is one defensible forecast that counts every layer a task touches, not a seat count that ignores consumption.
The line items to pin, the Flex Credit burn model, and the buyer side levers before you go agent first. Independent. Buyer side.
Join procurement and IT asset leaders who get our benchmarks on Salesforce, SAP, Oracle, Microsoft, and Workday AI pricing. No vendor spin.
