Agentforce Pricing: A Tier-by-Tier Breakdown

Agentforce is Salesforce’s platform for building and deploying AI agents across the Customer 360 ecosystem. It lets teams create both customer-facing agents (handling support conversations, answering product questions, or resolving cases) and user-facing agents (automating internal tasks, summarizing records, or assisting reps in the flow of work). Agents built on the platform can take autonomous actions, like updating a record or triggering a flow, rather than just generating text responses. It’s built to work natively within Salesforce data and processes, so agents draw on existing CRM context instead of operating as a separate bolted-on tool.

Salesforce has restructured Agentforce pricing around usage and business value instead of flat seat licenses. This guide breaks down each pricing tier, what it costs, and where the credits actually go. We’ll cover the free Foundations tier, the two usage-based options (Flex Credits and Conversations), and the flat-rate licensing paths built for employee-facing agents.

Tier 1: Salesforce Foundations (Free)

Salesforce Foundations gives every org a free entry point into Agentforce. We see this as the sandbox tier: it lets teams build and test before committing budget.

Foundations includes:

• Agentforce Builder
• Prompt Builder
• Agent Script
• Agentforce Coworker
• Agentforce Vibes

There’s no cost to start here. It’s the right tier for proof-of-concept work or for admins who want hands-on time with Agentforce before scaling up.

Tier 2: Flex Credits Explained ($500 per 100k credits)

Flex Credits are the pay-as-you-scale option for customer-facing and employee-facing agents. Salesforce prices this tier at $500 USD per 100,000 credits. Not every interaction with an agent draws credits. Only metered actions do. If an exchange doesn’t require a specific action, like updating a record or triggering a flow, there’s no charge for it.

This tier covers:

• Customer-facing agents
• Employee-facing agents
• Agentforce Voice
• Digital Wallet tracking
• Flexible buying models
• Agentforce Vibes

Salesforce introduced Flex Credits in 2025, shifting billing from a flat per-conversation rate to a per-action model. That year, Agentforce add-ons and Agentforce 1 Editions became generally available for Sales Cloud, Service Cloud, Field Service, and Industries Clouds, replacing the older Einstein add-ons and Einstein 1 Editions entirely. In other words, the tier structure this post walks through has only existed for about a year, and Salesforce has continued adjusting it since, most recently with a pay-per-resolution model for Help Agent in June 2026.

How Flex Credits Get Consumed

Each Agentforce Action draws credits from your pool. An action is any discrete task an agent performs: updating a record, summarizing a case, answering a product question, or running a custom flow.

Here’s the rate card for Agentforce usage:

Agentforce Actions

Usage Type	Production	Sandbox
Standard Action	20	16
Custom Action	20	16
Standard Voice Action	30	24
Custom Voice Action	30	24

Bring Your Own LLM: Connect External AI Models to Agentforce

BYOLLM lets customers connect an externally hosted large language model to Agentforce instead of relying only on Salesforce-managed models, extending the same choice Salesforce already offers for predictive AI. Setup happens in AI Models (formerly Einstein Studio), where admins add a foundation model from a supported provider, OpenAI, Microsoft Azure, Google Vertex AI, or Amazon Bedrock, and connect it with their own credentials. For anything outside those providers, including custom or in-house models, the LLM Open Connector allows a connection to any model that follows Salesforce’s Open Connector spec.

The external provider handles inference, but requests still route through the Models API and Einstein Trust Layer, so governance and data controls stay intact. Once connected, a BYOLLM model plugs into Prompt Builder and custom Agentforce actions like any Salesforce-managed model. On the credit side, Starter Prompts run 2 credits regardless of environment, so production and sandbox cost the same. That rate comes from the Flex Credits rate card you provided; I couldn’t independently confirm it elsewhere online.

Salesforce-Enabled Foundational LLMs

The table below breaks down the cost of using Salesforce’s own managed models, as opposed to models you connect yourself through BYOLLM. Each tier corresponds to a different level of model capability: Basic Prompts cost 2 credits, Standard Prompts cost 4 credits, and Advanced Prompts cost 16 credits, with the price rising as the underlying model gets more capable. These rates apply to Agentforce 1 Edition customers, Agentforce for Sales, Service, and Industries customers, and any other products that no longer bill through Einstein Requests. Orgs still on an Einstein Requests billing model wouldn’t use this multiplier structure.

For a specific project, it’s worth checking the Agentforce and Generative AI Usage and Billing help article for the latest guidance on what counts as “Basic” versus “Advanced.”

Usage Type	Multiplier
Basic Prompts	2
Standard Prompts	4
Advanced Prompts	16

Service and Speech Foundations

Beyond core Agentforce actions, two more usage types draw from the same Flex Credit pool. Help Agent Resolutions, tied to Salesforce’s Service offerings, run a flat 400 credits per resolution. Speech Foundations covers voice and language processing, and it’s priced differently depending on what the task involves. Converting speech to text costs by the hour of audio processed, while text-to-speech and translation scale by the volume of characters involved.

Usage Type	Unit	Multiplier
Speech to Text	Per hour of transcription	150
Text to Speech	Per 1M characters	6,000
Translation	Per 1M characters	4,000

Data 360 Pricing: How Volume-Based Credit Tiers Work

Data 360, formerly Data Cloud, is Salesforce’s unified data platform that powers Agentforce with customer context, pulling data from CRM records, warehouses, and unstructured sources into a single usable layer. Because it processes such large data volumes, Salesforce prices it differently than standard Agentforce Actions: usage scales with volume instead of a flat per-action rate.

Rates drop as monthly usage climbs through four tiers, from a base tier covering up to 300,000 credits through Tier 4, which kicks in after 12.5 million credits. Tiers reset on the first of each month and apply independently per usage type. Sandbox environments skip the tiered discount entirely and run at one fixed rate regardless of volume. A few examples from the production base tier:

Usage Type	Unit	Base Tier	Sandbox
Data 360 Prep	Per 1M rows	40	32
Data 360 Segmentation	Per 1M rows	50	40
Data 360 Activation	Per 1M rows	60	48
Data 360 Queries	Per 1M rows	3	2.4
Data 360 Unstructured Processing	Per 1MB	150	120

Tier thresholds reset on the first day of each calendar month. Customer 360 Platform usage types (Salesforce record operations and process invocation via Flows and Apex) are listed as TBA on the current rate card.

One note worth flagging for admins: sandbox multipliers apply to any pre-production environment, including scratch orgs, not just full sandboxes.

Tier 3: Conversations ($2 per conversation)

Conversations offer a simpler, per-conversation pricing model at $2 USD each. This tier is narrower in scope than Flex Credits:

• Customer-facing agents only
• Digital Wallet tracking included
• Pre-Purchase buying model only

The Conversations rate card is simple by comparison. Agentforce usage across ASA Messaging, Sales Coach, and SDR all carry a multiplier of 1 conversation credit. Salesforce notes that certain Agentforce features can also draw down Einstein Requests and Data Services Credits alongside conversations, so teams should watch more than one meter.

Employee-Facing Tiers: Add-Ons and Editions

For company-wide employee agent access, Salesforce offers licensing paths outside the credit-based model entirely.

Agentforce add-ons: $125 per user per month for Sales, Service, and Field Service. This includes unmetered Agentforce usage for employees plus Salesforce’s broader AI suite in the flow of work.

Agentforce Industries add-ons: $150 per user per month for Industries Clouds, layering industry-specific AI on top of everything in the standard add-ons.

Agentforce 1 Editions: start at $550 per user per month. This bundles the Agentforce add-on with 2.5M Flex Credits per org per year, aimed at orgs standardizing on Agentforce across Sales, Service, Field Service, and Industries.

Agentforce User License: $5 per user per month but requires a Flex Credits balance to function. It gives every employee limited CRM object access and meters usage against your Flex Credit pool rather than including it flat.

Agentforce Buying Models

Three purchase structures apply across these tiers:

• Pre-Purchase: Buy a set usage amount for the full contract term and pay upfront. This option saves the most, with usage drawing down from a prepaid balance.
• Pre-Commit: A committed usage arrangement.
• PayGo: Pay for usage as it happens.

All three models connect to Digital Wallet, which tracks Agentforce consumption in real time.

Conclusion

Agentforce pricing scales in three distinct directions: free experimentation through Foundations, metered growth through Flex Credits or Conversations, and flat-rate employee licensing through add-ons and Editions. The right tier depends on whether your agents are customer-facing or employee-facing, and whether your usage patterns are predictable enough to commit to a license versus pay-as-you-go credits.

Pricing details are subject to change. Confirm current rates with a Salesforce account executive before budgeting a rollout. Official Salesforce page can be found here.

Explore related content:

The Real Lessons Behind a Year of Agentforce Rollouts

Is Setup with Agentforce Ready for Prime Time?

Headless 360: Build an Agentforce Agent Using Natural Language