# Salesforce Agentforce — Salesforce's enterprise agentic-AI platform (rebranded "Agentforce 360") — prebuilt Service, SDR, Sales Coach and Commerce agents plus a low-code builder, running on live CRM data.

> Source: The Agents Index — https://theagentsindex.com/salesforce-agentforce (structured, researched, re-verified)
> Facts last verified: 2026-07-27

Salesforce Agentforce is Salesforce's enterprise agentic-AI platform — a suite of prebuilt AI agents (Service, SDR, Sales Coach, and several commerce agents) plus a low-code Agent Builder for creating custom ones, all running against a company's live Salesforce CRM data. It launched at Dreamforce in September 2024 and reached general availability on October 29, 2024; Salesforce reported that customers built more than 10,000 agents during the platform's Dreamforce launch week alone. Since 2025 Salesforce has folded its whole "Cloud" product line — Sales Cloud, Service Cloud, Marketing Cloud, Commerce Cloud — into a single rebrand called Agentforce 360, positioning agents rather than clouds as the organizing concept across its entire suite. At the center is the Atlas Reasoning Engine, which plans, executes and self-reflects on multi-step actions instead of following a fixed decision tree, alongside the Einstein Trust Layer, which masks sensitive data before it reaches a model. Salesforce itself is a public company (NYSE: CRM) founded in 1999 by Marc Benioff, Parker Harris, Dave Moellenhoff and Frank Dominguez, headquartered in San Francisco — this is an enterprise add-on to an existing CRM investment, not an independent startup product.

| Fact | Value |
| --- | --- |
| Website | https://www.salesforce.com/agentforce/ |
| Pricing | Foundations free ($0, 200K Flex Credits); Flex Credits $500/100K credits (~$0.10/action) or Conversations $2/conversation; User License ~$5/user/month + $125-$150/user/month add-ons; Agentforce 1 Editions from ~$550/user/month |
| API | Yes |
| Best for | Enterprises already running on Salesforce CRM that want agents acting directly on live sales/service/commerce data, with the budget and Salesforce admin expertise to manage a consumption-priced rollout. |
| Not for | Teams outside the Salesforce ecosystem, or buyers who want simple, predictable per-seat pricing rather than a metered Flex Credits/Conversations model layered on top of separate Data Cloud costs. |

## Pricing

| Tier | Price |
| --- | --- |
| Foundations | Free |
| Flex Credits | $500 / 100,000 credits |
| Conversations | $2 / customer-facing conversation |
| Agentforce User License | $5 / user/month |
| Agentforce 1 Editions | From ~$550 / user/month |

## Verdict

Agentforce is the CRM-native heavyweight of this category — no other general-purpose platform here puts agents this close to a company's actual sales, service and commerce data, and it backs that up with genuine model-agnostic support (GPT-5.x, Claude, Gemini) and a real developer surface (Agent API, Python SDK, CLI) rather than a walled-garden chatbot. The honest cost is Salesforce's own complexity: a Flex Credits/Conversations pricing model that has already reshaped itself multiple times since its October 2024 launch, a hard dependency on Data Cloud (now Data 360) for full cross-system context, and independent reports of multi-month rollouts and a real learning curve. Pick Agentforce if you're already deep in the Salesforce ecosystem and want agents acting directly on live CRM data; skip it if you want a lightweight platform you can stand up quickly outside that ecosystem, or predictable flat per-seat pricing.

## How it works

1. A team configures a prebuilt agent, such as Service, SDR, Sales Coach, or a commerce agent like Merchant, Buyer or Personal Shopper, or defines a custom one in the low-code Agent Builder using topics, instructions and actions, or Agent Script for tighter control.
2. The Atlas Reasoning Engine plans, executes and self-reflects on the multi-step actions needed to complete a request, instead of following a fixed decision tree.
3. The agent calls Salesforce Flows, Apex code and MuleSoft APIs to take action on CRM records directly, pulling in knowledge articles or outside data through Data Cloud for fuller context.
4. The Einstein Trust Layer masks sensitive data before it reaches the underlying model, which can be Salesforce's own Einstein models or, through the Models API and bring-your-own-model architecture, GPT-5.x, Claude or Gemini.

## Who it's for

- Enterprises already running Salesforce CRM that want agents acting directly on live sales, service and commerce data rather than a bolted-on chatbot
- Teams that want to start from prebuilt agents such as Service, SDR, Sales Coach or a commerce agent instead of building one from scratch
- Developers and admins who want either a low-code Agent Builder canvas or a full programmatic surface, including the Agent API, Python SDK and CLI, depending on how much control they need

## Strengths and weaknesses

- ✓ Deeper native CRM data and action integration than any other general-purpose platform in this category — agents call Salesforce Flows, Apex and MuleSoft APIs directly against live records, not just chat about them.
- ✓ Genuinely model-agnostic: supports OpenAI GPT-5.x, Anthropic Claude and Google Gemini models via its Models API and a documented bring-your-own-model (BYOM) architecture, alongside Salesforce's own Einstein models.
- ✓ Real production scale and a genuine developer surface: general availability since October 29, 2024, Salesforce reported customers built 10,000+ agents during Agentforce's Dreamforce launch week alone, and there's a documented Agent API, Python SDK, Mobile SDK and CLI for building/testing outside the point-and-click UI.
- ✗ Consumption pricing that has already changed shape multiple times since its October 2024 launch — Flex Credits and Conversations are mutually exclusive within one org, per-user licenses still consume credits on top of a monthly fee, and independent buyer analyses put real mid-market annual spend at roughly $150,000-$600,000 once Data Cloud (Data 360) implementation is included.
- ✗ Full cross-system context requires Data Cloud (rebranded Data 360) as a separate, additionally-priced dependency — an agent limited to core Salesforce objects alone sees a narrower slice of a company's data.
- ✗ Independent buyer reports describe a real setup burden: pilots commonly take 4-6 weeks and multi-agent production rollouts 3-6 months, with users citing a clunky configuration UI and a steep learning curve, especially across multiple service desks.
- ⚠ Requires an existing Salesforce investment to get full value — the prebuilt agents assume Sales/Service/Commerce Cloud (now Agentforce-branded) data underneath them.
- ⚠ Consumption-based pricing across multiple mutually-exclusive credit models makes total cost genuinely hard to forecast without a vendor-guided estimate.

## Key features

- **Atlas Reasoning Engine** — Plans, executes and self-reflects on multi-step actions instead of following a fixed decision tree.
- **Prebuilt vertical agents** — Service, SDR, Sales Coach, Merchant, Buyer and Personal Shopper agents ready to configure, not build from scratch.
- **Agent Builder + Agent Script** — A low-code canvas for defining topics/instructions/actions, with a hybrid natural-language-plus-rules language for tighter control.
- **Model-agnostic (BYOM)** — Runs OpenAI, Anthropic and Google models via the Models API, on top of Salesforce's own Einstein models.
- **Real developer surface** — Agent API, Python SDK, Mobile SDK and a CLI (Agentforce DX) for building and testing agents outside the UI.
- **Einstein Trust Layer** — Masks sensitive CRM data before it reaches a model — an enterprise compliance layer most smaller platforms don't ship.

## Use cases

- **24/7 tier-1 support deflection** — A support org configures the Service Agent against its knowledge base and Service Cloud case data so routine tickets resolve without a human, with handoff for anything it can't close.
- **Always-on SDR qualification** — A sales team runs the SDR Agent to engage inbound leads and answer objections around the clock, only routing a rep in once a prospect is ready to book.
- **A bespoke internal agent against live CRM data** — An admin uses Agent Builder (or a developer uses Agent Script/the CLI) to define a custom agent that takes action directly on Salesforce records — something the prebuilt agents don't cover.

## Integrations

Salesforce Flows, Apex and MuleSoft APIs (native) · Data Cloud / Data 360 (cross-system context) · Slack · Models API — OpenAI, Anthropic, Google + BYOM · Agent API, Python SDK, Mobile SDK (iOS/Android), Agentforce DX (CLI + VS Code) · Model Context Protocol (MCP) — a native client connects to any MCP-compliant server, plus verified servers from AgentExchange partners

## Sources

- https://www.salesforce.com/news/press-releases/2024/10/29/agentforce-general-availability-announcement/
- https://www.salesforce.com/news/stories/agentforce-launch-zone-announcement/
- https://developer.salesforce.com/docs/ai/agentforce/guide/get-started-agents.html
- https://developer.salesforce.com/docs/ai/agentforce/guide/supported-models.html
- https://www.convopro.io/blog/salesforce-agentforce-pricing-in-2026-how-the-models-work-and-how-to-choose
- https://www.default.com/post/salesforce-agentforce-review-and-pricing
- https://www.oliv.ai/blog/salesforce-agentforce-reviews-analyzed
- https://www.salesforce.com/agentforce/mcp-support/