Agentforce
Salesforce · Salesforce's AI Agent Platform for Service, Sales, and Commerce
Agentforce is Salesforce's platform for building and deploying autonomous AI agents grounded in Salesforce CRM and Data 360, with ready-made agents for customer service, IT and HR requests, shopping, sales pipeline work, and supply chain. It is aimed at companies already running on Salesforce that want agents working inside their existing customer data, workflows, and channels.
Best for
Companies whose customer data, service processes, and sales workflows already run on Salesforce and that want agents acting directly on that data across service, sales, commerce, and employee support
Not ideal for
Teams not on Salesforce, small businesses wanting a simple self-serve support bot with one flat price, or organizations that need to self-host or move agents between platforms
Who it's for
Salesforce customers in service, sales, commerce, IT, and HR who want autonomous agents grounded in their CRM data
Capabilities
- Agentforce Builder: a doc-style canvas for writing agents in natural language combined with Agent Script, with subagents, actions, variables, and if-then logic
- Actions built from Salesforce Flows, Apex, MuleSoft APIs, prompt templates, and knowledge articles, plus external systems
- Grounding in Data 360 and Customer 360 CRM data, with the Einstein Trust Layer providing dynamic grounding, zero data retention, and toxicity detection
- Job-ready agents: Casey (customer service), Paige (IT and HR), Carter (shopping), Piper (inbound leads), Marshall (supply chain), and Fin, all generally available
- Long-horizon runtime for agents that pursue goals over days and weeks, with memory, durable execution, and dynamic steering (first used by the Hunter outbound sales agent)
- Deployment across web, phone, and apps, including WhatsApp, Messenger, SMS, voice, and Slack
- Simulator and batch testing that turns real interactions into reusable test cases, plus analytics for monitoring agent performance
- Agent Script toolchain (parser, linter, compiler, language server, VS Code extension) released as open source under Apache 2.0
- Agent API, a REST API for starting sessions and exchanging messages with agents from external apps, authenticated through an OAuth external client app
Limitations
- Built around the Salesforce platform: agents are configured in Salesforce, grounded in Salesforce data, and sold as Salesforce add-ons, editions, or credits
- The agent runtime is proprietary: Agent Script compiles to a Salesforce-internal specification, so agents only run on Salesforce infrastructure
- Several pricing models (Flex Credits per action, per-conversation, per-user add-ons, Agentforce 1 Editions) make total cost hard to estimate without modeling your volume
- Hunter, the first long-horizon agent, is still in pilot (GA planned for November 2026), and building your own long-horizon agents is described as a future capability
- Custom actions often rely on Flows, Apex, or MuleSoft, which assumes Salesforce admin or developer skills
- No self-hosted deployment is documented
Use cases
- Deploying Casey to resolve order-status and account questions over web chat, WhatsApp, SMS, and voice, with escalation to human agents
- Handling employee IT and HR requests in Slack and internal portals with Paige
- Adding Carter to an online store so shoppers can compare products and check out inside the chat
- Qualifying inbound website and email leads with Piper before handing them to sales reps
- Building a custom service agent in Agentforce Builder that calls existing Flows and Apex classes to process returns
Agentforce pricing
Salesforce Foundations
$0
- Agentforce Builder, Prompt Builder, and Agent Script
- Agentforce Coworker and Agentforce Vibes
- $0 add-on for Salesforce Enterprise Edition and higher
Flex Credits
$500 per 100,000 credits
- Standard Agentforce action: 20 credits
- Agentforce Voice action: 30 credits
- Pre-purchase, pre-commit, or pay-as-you-go buying options
Conversations
$2 per conversation
- For customer-facing agents only
Agentforce User License
$5/user/month
Billed monthly
- Requires a separate Flex Credits purchase
Agentforce add-ons (Sales, Service, Field Service)
$125/user/month
Billed monthly
- Unmetered usage
- Industries add-ons are $150/user/month
Agentforce 1 Editions
From $550/user/month
Billed monthly
- Agentforce add-on included
- 2.5M Flex Credits per org per year
Note: Salesforce sells Agentforce through several models that can be combined: consumption (Flex Credits or per-conversation), per-user add-ons, and editions. Credits can be pre-purchased upfront, pre-committed with monthly billing (Salesforce says widely available later this year), or paid as you go and billed monthly in arrears. Salesforce notes that actual credit usage varies with activity levels and features used.
Technical specs
Agentforce vs. competitors
Agentforce vs. Fin
Fin is now owned by Salesforce and appears in Agentforce's lineup, but Salesforce positions it as a fast-to-deploy agent that works with the help desks companies already use; Agentforce is the broader platform for building agents across service, sales, commerce, and employee support on Salesforce data.
Agentforce vs. Sierra
Sierra is a standalone platform for branded customer-facing agents and is not tied to one CRM; Agentforce is strongest when your customer data and processes already live in Salesforce.
Agentforce vs. Zendesk AI Agents
Zendesk AI Agents are built into Zendesk's ticketing and routing; Agentforce plays the same role for Salesforce Service Cloud and extends to sales, commerce, IT, and HR agents.
Agentforce vs. Decagon
Decagon focuses on enterprise support across chat, voice, email, and SMS; Agentforce covers support plus sales, commerce, and back-office agents inside the Salesforce platform.
Frequently asked questions
What is Agentforce?
Agentforce is Salesforce's platform for building and running autonomous AI agents. Agents are grounded in Salesforce CRM and Data 360, take actions through Flows, Apex, and MuleSoft, and work across web, phone, messaging apps, and Slack. Salesforce also sells ready-made agents for service, IT and HR, shopping, sales, and supply chain.
How much does Agentforce cost?
Salesforce lists several models: Flex Credits at $500 per 100,000 credits (a standard action uses 20 credits, a voice action 30), $2 per conversation for customer-facing agents, Agentforce add-ons for Sales, Service, or Field Service at $125/user/month ($150 for Industries), Agentforce 1 Editions from $550/user/month, and a $5/user/month user license paired with Flex Credits. Salesforce Foundations, a $0 add-on for Enterprise Edition and higher, is the free starting point.
Do I need Salesforce to use Agentforce?
Agentforce is built on the Salesforce platform: agents are configured in Agentforce Builder, grounded in Salesforce data, and priced as Salesforce add-ons, editions, or credits. Salesforce says every Salesforce customer can start with Agentforce for free through Salesforce Foundations, a $0 add-on for Enterprise Edition and higher.
What are Agentforce's job-ready agents?
Announced in September 2026, they are Casey (customer service across voice, SMS, WhatsApp, and web chat), Paige (IT and HR requests), Carter (shopping and in-chat checkout), Piper (inbound lead qualification), Marshall (supply chain), and Fin, all generally available, plus Hunter (outbound sales), in pilot with GA planned for November 2026.
Does Agentforce have an API?
Yes. The Agent API is a REST API for starting sessions, sending messages, and ending sessions with Agentforce agents from outside Salesforce. It requires an external client app with OAuth client credentials, does not support agents of the type "Agentforce (Default)", and its usage counts toward Agentforce credit consumption.
How does Agentforce relate to Fin?
Salesforce completed its acquisition of Fin, formerly Intercom, on September 10, 2026. Fin appears in Agentforce's job-ready agent lineup, and Salesforce says it continues to work with the help desks companies already use. Fin has its own listing in this directory.
Integrations & fit
Alternatives to consider
Sierra
Enterprises that want branded, action-taking AI agents as the front door to their customer experience, with deep control over tone and guardrails
CustomDecagon
Large consumer brands automating high-volume, multi-channel customer support with on-brand, measurable AI agents
Custom
Zendesk
Support teams already on Zendesk who want AI-powered ticket resolution and triage across channels
PaidAda
Mid-market and enterprise teams that want a vendor-independent AI support platform working across all channels
CustomAbout Agentforce
Agentforce is less a single bot than the agent layer of the Salesforce platform. You build agents in Agentforce Builder, a doc-style canvas where agents are written in natural language combined with Agent Script, Salesforce's language that mixes natural-language instructions with programmatic logic. Agents act through the same building blocks Salesforce admins already use: Flows, Apex, MuleSoft APIs, prompt templates, and knowledge articles. They are grounded in Data 360 and Customer 360, run on the Atlas Reasoning Engine, and pass through the Einstein Trust Layer, whose guardrails include dynamic grounding, zero data retention, and toxicity detection. Agents work across web, phone, and apps, with testing tools that turn real interactions into reusable batch tests. In September 2026 Salesforce also started selling named, job-ready agents: Casey for customer service across voice, SMS, WhatsApp, and web chat, Paige for IT and HR requests, Carter for shopping and in-chat checkout, Piper for inbound leads, Marshall for supply chain, and Fin, which Salesforce acquired on September 10, 2026 and which still works with other help desks. Hunter, an outbound sales agent in pilot with general availability planned for November 2026, is the first agent on a new long-horizon runtime that pursues goals over days and weeks. Pricing comes in several forms: Flex Credits billed per action, $2 per conversation for customer-facing agents, per-user add-ons, and Agentforce 1 Editions, with a $0 starting point through Salesforce Foundations. The tradeoffs follow from that design: Agentforce is strongest when your data and processes already live in Salesforce, its runtime is proprietary to Salesforce infrastructure, and choosing among the pricing models takes real modeling of expected volume.
Updates from Agentforce
Salesforce introduced a portfolio of named agents: Casey (help), Paige (IT and HR), Carter (shopper), Piper (inbound pipeline), Marshall (supply chain), and Fin, all generally available, plus Hunter (outbound sales) in pilot with GA planned for November 2026. Hunter is the first agent on a new long-horizon runtime that pursues goals across days and weeks.
Salesforce closed its acquisition of Fin, the customer agent platform formerly known as Intercom, adding it to Salesforce's customer service AI portfolio. Salesforce says Fin continues to work with the help desks and systems companies already use.
Are you the founder? Claim this listing →