Salesforce Launches Agentforce 3 to Boost Visibility, Control, and Agent Performance

SAN FRANCISCO, June 23, 2025 — Salesforce today announced Agentforce 3: a major upgrade to its digital labor platform that gives companies the visibility and control to scale AI agents without compromise. As enterprise adoption accelerates, the real blocker has become clear: teams can’t see what agents are doing — or evolve them fast enough. Agentforce 3 changes that.

Built on learnings from thousands of Agentforce deployments since its initial launch in October 2024, Agentforce has helped customers deliver undeniable value. This includes reducing Engine’s average customer case handle time by 15%, autonomously resolving 70% of 1-800Accountant’s administrative chat engagements during critical tax weeks in 2025, and increasing Grupo Globo’s subscriber retention by 22%. Agentforce 3 equips leaders to monitor, improve, and scale their AI workforce with confidence.

With a new Command Center for complete observability, built-in support for Model Context Protocol (MCP) for plug-and-play interoperability, and over 100 new prebuilt industry actions to speed time to value, Agentforce 3 helps companies scale what works, fix what doesn’t, and unlock the full potential of agentic AI — with clarity, control, and speed.

AI agent adoption is surging. According to a soon to be released Slack Workflow Index, AI agent usage is up 233% in six months, and over that same period, 8,000 customers have signed up to deploy Agentforce. But until now, agent platforms have lacked the tooling, governance, and observability needed to scale enterprise-wide. Agentforce 3 closes this gap — delivering the complete visibility, secure tool integration, and enterprise-grade controls organizations need to make agent velocity their competitive advantage.

“With Agentforce, we’ve unified agents, data, apps, and metadata to create a digital labor platform, helping thousands of companies realize the promise of agentic AI today,” said Adam Evans, EVP & GM of Salesforce AI. “Over the past several months we’ve listened deeply to our customers and continued our rapid pace of technology innovation. The result is Agentforce 3, a major leap forward for our platform that brings greater intelligence, higher performance, and more trust and accountability to every Agentforce deployment. Agentforce 3 will redefine how humans and AI agents work together — driving breakthrough levels of productivity, efficiency, and business transformation.”

Agentforce Command Center

As AI agents take on routine tasks and begin collaborating more closely with human teammates, teams need a new observability layer built for the era of digital labor. Agentforce Command Center is that layer: a complete observability solution that gives leaders a unified pane of glass to monitor agent health, measure performance, and optimize outcomes. Built into Agentforce Studio, it completes the agent lifecycle with powerful tools to understand and refine agents at scale.

• Uncover patterns across interactions to optimize your agents: Command Center empowers teams to analyze every AI agent interaction, drill into specific moments, understand trends in usage, and see AI-powered recommendations for tagged conversation types to continuously improve your Agentforce.
• Track agent health and intervene in real time: Get live, detailed analytics for latency, escalation frequency, and error rates, plus real-time alerts when the unexpected happens, so teams can act fast and keep agents running smoothly.
• Understand what’s working, and where to improve: Command Center offers detailed dashboards that track agent adoption, feedback, success rates, cost, and topic performance — so teams can see what’s gaining traction and where to improve.
• See what your agents are doing — in the tools your teams already use: Agentforce captures all agent activity in a native, extensible session-tracing data model in Data Cloud — powering analytics, monitoring, and real-time alerting. Built on the OpenTelemetry standard, these agent signals integrate seamlessly with tools your teams already use, including Datadog, Splunk, Wayfound, and other monitoring partners for end-to-end visibility across your existing stack.
• Deliver a configurable Command Center for every team: Monitor AI agents alongside human teammates — right in the flow of work. Starting with Service Cloud, agent activity will surface in real-time wallboards so contact center supervisors can track performance and escalate fast. And over time, every department will have a Command Center purpose-built for optimizing their agents.
• Build and test agents fast with AI-assisted development tools: In Agentforce Studio, use natural language to generate topics, instructions, and test cases. Testing Center simulates behavior at scale with state injection and AI-driven evals — so you can pressure-test your agents before going live.

Enabling Secure Enterprise Connectivity with MCP and A2A Support

AI agents can’t drive impact if they can’t take action using the tools your business relies on. As open standards like Model Context Protocol (MCP) gain traction, they bring new opportunities for interoperability, but also challenges around governance, identity, and control. Agentforce 3 solves this by pairing open connectivity with enterprise-grade trust — giving agents native access to the tools they need, without compromising on control.

• MCP support built natively into Agentforce: Agentforce will include a native MCP client, enabling Agentforce agents to connect to any MCP-compliant server — no custom code required. Like a “USB-C for AI,” this enables access to enterprise tools, prompts, and resources — governed by your existing security policies.
• Turn APIs into MCP servers instantly with MuleSoft: Leveraging new MCP connectors, MuleSoft converts any API and integration into an agent-ready asset, complete with security policies, activity tracing, and traffic controls — empowering teams to orchestrate and govern multi-agent workflows.
• Easily host and manage custom MCP servers with Heroku: Heroku Managed Inference and AppLink make it fast and easy to deploy, register, maintain, and connect your custom MCP servers. With Heroku’s secure infrastructure and DevOps automation, developers can bring trusted custom actions to Agentforce with less friction.

Enhancing the Agentforce Architecture for Unmatched Enterprise Readiness

Underpinning every new capability in Agentforce 3 is an enhanced Atlas architecture, providing an enterprise-ready foundation through lower latency, greater accuracy, global availability, and additional options for control through new LLMs hosted on Salesforce infrastructure.

• Expanded LLM choice with hosted Anthropic: Agentforce can now use Anthropic’s Claude Sonnet model hosted via Amazon Bedrock within the Salesforce trust boundary to meet the needs of customers in high compliance industries. As part of this expanded relationship, Anthropic will work with Salesforce to empower customers in regulated industries to scale Agentforce adoption with Claude. Later this year, Salesforce will also allow customers to use Google’s Gemini in Agentforce, solidifying Agentforce’s position as the leader in trusted, flexible AI agents.
• Faster performance and response streaming: Experience a more responsive Agentforce with 50% lower latency since January 2025. Response streaming is also generally available in this release, so users can see answers appear in real time.
• Greater accuracy through web search, inline citations: Agentforce 3 features trusted, expanded grounding through the addition of web search as a data source, allowing agents to go beyond internal data to answer requests as well as inline citations that provide references to the grounding sources used in responses.
• More languages, more geographies: Agentforce 3 features an expanded global footprint, deploying to Canada, the U.K., India, Japan, and Brazil to serve AI agent traffic within those regions. This release also adds GA support for six new languages, including French, Italian, German, Spanish, Japanese, and Portuguese — with more than 30 additional languages rolling out in the coming months.
• Enhanced resiliency with automatic model failover: To ensure agents are always on, Agentforce now enables automatic, latency-based failover — dynamically shifting traffic between model providers in case of performance degradation or outages.
• Bringing Agentforce to Public Sector customers with FedRAMP High Authorization: Agentforce is now authorized and generally available in Government Cloud Plus, which enables public sector customers to bring Agentforce to their missions with the highest standards of trust, security, and compliance.

Available Today:

• Agentforce 3
• Agentforce adoption analytics
• Testing Center enhancements
• 100+ new, pre-built industry actions
• New Agentforce add-on SKUs with unlimited employee action usage
• Heroku managed MCP server hosting
• Increased speed and response streaming
• Web Search for Agentforce Data Libraries
• Agentforce for Government Cloud Plus with FedRAMP High authorization
• Expanded global availability (Canada, U.K., India, Japan, Brazil) and language support (French, Italian, German, Spanish, Japanese, Portuguese)

In Pilot or Beta Today:

• Anthropic Claude models hosted within the Salesforce trust boundary — generally available in July
• MuleSoft MCP and A2A support — generally available in July
• Heroku AppLink — generally available in July
• Session Tracing Data Model — generally available in August
• Agent health monitoring — generally available in August

To Be Released Soon:

• Agentforce native MCP support – July
• Agentforce Command Center and Agentforce Studio app – August

About Salesforce

Salesforce helps organizations of any size reimagine their business for the world of AI. With Agentforce, Salesforce’s trusted platform, organizations can bring humans together with agents to drive customer success—powered by AI, data, and action. Visit www.salesforce.com for more information.

Source: Salesforce