RXed AI News

We x-ray the AI industry.
@RXed_EU
SCORE 7/1019/20 elements Audited 2026-07-31 · RXed table v1.0

Salesforce Agentforce

“Build, customize, test, deploy, and monitor AI-powered agents” — the vendor’s own words

The broadest enterprise agent platform on paper — Atlas, MCP, A2A, real observability — but $800M ARR and CIOs saying 'it isn't there yet' are both true; your agents will only be as good as your Data Cloud.

Enterprise platformAutomation & AgentsSalesChatPaid
Vendor
Salesforce · www.salesforce.com
Origin
US — San Francisco
Pricing
Salesforce Foundations Free · Flex Credits $500 per 100k credits · Conversations $2 per conversation · Agentforce add-ons $125/user/mo · Agentforce 1 Editions from $550/user/mo · Agentforce User License $5/user/mo
Users (official only)
29,000+ Agentforce deals closed since launch; $800M Agentforce ARR, up 169% Y/Y (source, 2026-02-25)
Salesforce FoundationsFreeAgentforce Builder, Prompt Builder, Agent Script, Coworker, Vibes
Flex Credits$500 per 100k credits€500 official EUR list; standard action 20 credits, Voice action 30 credits
Conversations$2 per conversation€2 official EUR list; customer-facing agents, pre-purchase only
Agentforce add-ons$125/user/mounmetered employee agent usage; Industries add-ons $150
Agentforce 1 Editionsfrom $550/user/moincludes add-on + 2.5M Flex Credits per org per year
Agentforce User License$5/user/mocompany-wide employee access; requires Flex Credits

Unused Flex Credits do not roll over; no overage penalty (contracted rate billed in arrears). Examples exclude Data 360 consumption, which grounding-quality effectively requires. Analysts flag consumption unpredictability as a mid-market budgeting problem.

checked 2026-07-31 · vendor pricing page

Element scores

Reactive
Retrieval & Memory
Orchestration
Validation
Models
Primitives
Pr8
Prompts
Em7
Embeddings
Cx8
Context
Tr8.5
Tracing
Lg7.5
LLM
Compositions
Fc8
Function calling
Vx7
Vector store
Rg7.5
RAG
Gr6.5
Guardrails
Mm7
Multimodal
Deployment
Ag7
Agents
Ft3
Fine-tuning
Fw8.5
Frameworks & harnesses
Ev7
Evaluations
Sm4
Small models
Emerging
Ma7.5
Multi-agent
Sy4.5
Synthetic data
Pc8.5
Protocols
In
Interpretability
Th7.5
Thinking models
Tap or hover any element to see why it got that score.

Strengths

No other enterprise platform covers this much of the periodic table: configurable Atlas hybrid reasoning with real model choice (OpenAI, Claude in the trust boundary, Gemini, xGen), Command Center observability on OpenTelemetry, a Testing Center with synthetic-utterance batch evals, Agentforce Voice, and the best protocol posture in the category — native MCP client, 35+ partner MCP servers on AgentExchange, MuleSoft API-to-MCP conversion, and co-creatorship of A2A. Official EUR price list published. FedRAMP High for US federal.

Honest dings

Two publicly disclosed prompt-injection chains inside a year (ForcedLeak CVSS 9.4, then PipeLeak) — handled responsibly, but proof the attack surface is real. The July 2026 KeyBanc CIO survey found customers unwilling to pay CRM-provider AI premiums and calling the product immature, and Forrester flags complex pricing and slow value realization. Credits don't roll over, mid-market cost predictability is poor, there's no customer fine-tuning, and agent quality collapses on messy data — Data Cloud is a de facto prerequisite with its own bill.

Best for: Salesforce-first enterprises with clean Data Cloud foundations that want governed, observable agents inside their CRM perimeter — not mid-market teams hoping to bolt AI onto messy data at a predictable monthly cost.
Visit salesforce.com/agentforce Prices and details change — this passport is re-verified at least quarterly.

Sources

Every audit lists the research it rests on — transparency and traceability are the product. Tools evolve: each audit is a snapshot of its audit date, and re-audits supersede older versions (kept below for reference).