Salesforce Launches Agentforce 3.0: Here’s Everything You Need to Know About the Upgrade

What do Pokémon and Agentforce have in common?

They evolve when it’s time to level up!

Just like Pikachu becomes Raichu to gain more power, Agentforce has transformed into its next evolution — Agentforce 3.0!

All hail the all-new Agentforce 3.0!

On June 23, 2025, Salesforce officially announced the launch of Agentforce 3.0, marking a major leap forward from its previous versions. And this upgrade isn’t just about bells and whistles – it’s about listening to the ecosystem.

With enhanced visibility, stronger governance, and seamless integrations, Agentforce 3.0 is designed to combat challenges that Salesforce professionals have been talking about — from concerns about security and technical debt to pricing, governance, and ROI.

So, what’s new? And how can you leverage this powerful update?

Let’s break it down, feature by feature, and see how Agentforce 3.0 is helping teams unlock the true potential of AI-powered agents.

Just In: Salesforce Unveils Agentforce 3.0 — Here’s What’s New in the Latest Update

In just six months, AI agent adoption has increased by 233%, and 8,000+ businesses have already embraced Agentforce to boost productivity and streamline operations⁽ⁱ⁾.

But scaling AI across the enterprise hasn’t been easy, until now.

Agentforce 3.0 is here to change that. The latest update bridges the critical gaps that have long held businesses back: visibility, security, and governance. With a robust architecture designed for today’s enterprise needs, Agentforce 3.0 turns agent velocity into a true competitive edge.

Here’s what’s new in Agentforce 3.0:

Built for Open Standards and Plug-and-Play Interoperability

Agentforce 3.0 isn’t just smarter — it’s more connected. With native support for Model Context Protocol (MCP) and an expanded AgentExchange, businesses can now seamlessly integrate with 30+ top-tier partners like AWS, Google Cloud, IBM, Stripe, Notion, PayPal, and many more. No more silos, just smooth interoperability across the tools you already use.

Here’s what MCP empowers you to do:

• Turn APIs into agent-ready capabilities with MuleSoft
  New MCP connectors in MuleSoft allow you to instantly convert any API or integration into an MCP-compatible service. These services come equipped with built-in governance, like security, observability, and traffic controls, so you can easily design and run powerful multi-agent workflows across your entire ecosystem.

• Deploy custom MCP services with Heroku AppLink
  Developers can now use Heroku to quickly stand up custom MCP services. This offers a flexible, secure, and scalable way to connect your internal tools and workflows to Agentforce, leveraging the automation and simplicity Heroku is known for.

• Empower secure agent interactions in Slack
  Slack will soon introduce its own MCP server, built in collaboration with Anthropic. This will allow Agentforce agents to interact with Slack messages, files, and threads – delivering actionable insights and triggering workflows, while staying within strict enterprise security and permission boundaries.

• Centralize control with the Agentforce Gateway
  Built by MuleSoft, the unified Agentforce gateway acts as a central command hub. It manages agent identity, connection permissions, and policy enforcement, giving IT full control over which agents can access which services, how they behave, and ensuring enterprise-grade compliance every step of the way.

Powered by the Upgraded Atlas Architecture

At the core of Agentforce 3.0 is a reengineered Atlas architecture, delivering performance that’s fast, accurate, and resilient. It supports natively hosted large language models (LLMs), including Anthropic, offering the flexibility and control that enterprises demand from AI systems.

100+ New Industry Actions to Accelerate Value

Whether you’re in finance, healthcare, retail, or beyond, Agentforce 3.0 includes over 100 prebuilt industry actions, making it easier than ever to customize agents for your unique business processes and see results faster.

Meet the New Agentforce Command Center

As AI agents become vital digital coworkers, handling repetitive tasks and collaborating in real-time, leaders need a smarter way to manage them. That’s where the Agentforce Command Center comes in.

This powerful observability hub, integrated directly into Agentforce Studio, gives organizations a unified view to monitor agent performance, ensure operational health, and fine-tune outcomes at scale. It’s built for the age of AI.

Here’s how you can optimize, test, and scale with the all-new Agentforce Command Center.

• Spot trends and fine-tune agent performance
  The Command Center allows teams to explore patterns across every AI agent interaction. By drilling down into specific conversation types, understanding usage trends, and acting on intelligent recommendations, businesses can continuously refine their Agentforce deployment for better outcomes.

• Stay ahead with real-time health insights
  Keep tabs on agent health with live analytics that track latency, error rates, and escalation trends. Get instant alerts when something goes off track — empowering teams to step in and resolve issues before they impact the user experience.

• Identify what’s working and what’s not
  From adoption rates to feedback, topic performance, cost, and success metrics, the Command Center delivers in-depth dashboards that reveal how agents are performing and where improvements are needed.

• Get end-to-end visibility with your existing tools
  Agent activity is captured in a native session-tracing model powered by Data Cloud, built on OpenTelemetry standards. This makes it easy to integrate with platforms your teams already rely on—like Datadog, Splunk, Wayfound, and more, ensuring seamless monitoring across your stack.

• Custom Command Centers for every team
  Starting with Service Cloud, Agentforce enables teams to track agent and human collaboration side-by-side. Real-time wallboards keep supervisors in the loop, enabling fast escalations and performance insights. And as the platform evolves, each department will get a tailored Command Center to meet their unique needs.

• Accelerate development with AI-powered tools
  Within Agentforce Studio, teams can now build and test agents faster using natural language inputs. The built-in Testing Center simulates scenarios at scale using state injection and AI-based evaluations, so agents are performance-ready before they go live.

Agentforce 3.0’s Powerhouse: The Enhanced Atlas Reasoning Engine

Agentforce 3.0 is built on a reimagined Atlas architecture, engineered to meet the demands of large-scale enterprise deployment. With faster response times, accurate outputs, broader global availability, and greater control over AI model selection, this upgraded foundation ensures Agentforce is ready for anything your business throws at it.

Here’s what’s new in the upgraded Atlas architecture:

More Choice, More Control With Hosted LLMs

Now supporting Anthropic’s Claude Sonnet hosted securely within Salesforce’s trust boundary (via Amazon Bedrock), Agentforce offers trusted AI options tailored for highly regulated industries. This collaboration helps businesses in sectors like healthcare, finance, and public services scale confidently.

Also, coming soon: support for Google’s Gemini — further expanding your model flexibility.

Lightning-Fast Responses, Real-Time Streaming

Agentforce 3.0 delivers 50% lower latency compared to earlier this year⁽ⁱⁱ⁾, ensuring lightning-fast answers. Plus, with response streaming now generally available, users can watch replies unfold in real-time.

Sharper Accuracy With Real-World Context

Agentforce now taps into live web search for deeper grounding and delivers inline citations for added transparency. Your agents don’t just generate responses, they back them up with trusted sources.

Global Reach, Local Expertise

Agentforce is expanding its global footprint, now routing AI traffic through new regions including Canada, the U.K., India, Japan, and Brazil. Alongside that, it now supports six additional languages – French, Italian, German, Spanish, Japanese, and Portuguese – with 30+ more rolling out soon. Wherever your teams are, Agentforce is ready.

Built-In Resilience for Uninterrupted Performance

When every second counts, downtime isn’t an option. That’s why Agentforce now features automatic failover, dynamically shifting traffic to alternate model providers if performance dips, so your agents stay responsive, always.

Trusted AI for the Public Sector

Agentforce has officially earned FedRAMP High Authorization and is now available in Government Cloud Plus. Public sector businesses can now harness AI agents that meet the highest benchmarks of trust, compliance, and security.

Industry-Ready Out of the Box

Agentforce 3.0 packs over 200 prebuilt industry actions—half of which are brand new – designed to accelerate outcomes across use cases like patient scheduling, advertising proposals, vehicle service management, and more. Paired with simplified, per-user pricing for Sales, Service, and Industry Cloud editions (with unlimited usage for internal agents), it’s never been easier to start strong and scale fast.

What’s Available, What’s Coming: Your Agentforce 3.0 Rollout Guide

Available Now:

• Agentforce 3.0 – Enterprise-grade AI agents, ready for action.
• Adoption Analytics – Track usage trends and measure success across teams.
• Enhanced Testing Center – Build, simulate, and refine agents at scale.
• 100+ New Prebuilt Industry Actions – Accelerate value with ready-to-use workflows.
• New Agentforce Add-On SKUs – Includes unlimited usage for employee-facing agents.
• Managed MCP Hosting on Heroku – Easily deploy custom services with built-in automation.
• Faster Performance & Response Streaming – Answers appear in real-time with up to 50% lower latency.
• Web Search for Agentforce Data Libraries – Expand knowledge grounding beyond internal data.
• Agentforce for Government Cloud Plus – FedRAMP High authorized for secure public sector use.
• Expanded Global Reach – Now supporting deployments in Canada, the U.K., India, Japan, and Brazil.
• New Language Support – French, Italian, German, Spanish, Japanese, and Portuguese, with more on the way.

In Pilot or Beta:

• Hosted Anthropic Claude Models – Running securely within the Salesforce trust boundary (GA in July).
• MuleSoft MCP & Agent-to-Agent (A2A) Support – Bringing seamless service orchestration (GA in July).
• Heroku AppLink – Fast-track your custom MCP services (GA in July).
• Session Tracing Data Model – Fine-grained observability powered by Data Cloud (GA in August).
• Agent Health Monitoring – Real-time insights to keep agents performing optimally (GA in August).

Coming Soon:

• Agentforce Native MCP Support – Unlock richer agent interoperability (Launching July).
• Agentforce Command Center & Studio App – Full lifecycle management in one place (Arriving August).

Agentforce 2.0 vs Agentforce 3.0: Differences at a Glance

While Agentforce 2.0 brought the vision of digital labor to life, Agentforce 3.0 is where it truly scales.

Let’s break down the key differences between Agentforce 2.0 and 3.0.

Conclusion

Just six months into the year, and Salesforce has already rolled out multiple iterations of Agentforce, each one pushing the boundaries of what AI agents can do.

With the launch of Agentforce 3.0, the platform addresses one of the biggest roadblocks to enterprise-wide adoption: the need for greater visibility, agility, and control. Backed by real-world impact — from reducing case handling times to driving customer retention — Agentforce continues to prove its value across industries and use cases.

From reducing case resolution times to autonomously managing high-stakes tax week queries, the results speak for themselves. And now, with a reimagined architecture, enhanced controls, and a purpose-built Command Center, teams finally have the tools to scale confidently.

Truly, Agentforce 3.0 redefines digital work, setting the stage for AI agents to become indispensable partners in every enterprise.

Statistics References:

^{(i) & (ii)} Salesforce