Salesforce Launches Agentforce 3 to Solve Biggest Blockers to Scaling AI Agents: Visibility and Control

Salesforce, the world’s no. 1 Artificial Intelligence (AI) CRM, has announced Agentforce 3, a major upgrade to its digital labour 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 autonomously resolving 70% of 1-800Accountant’s administrative chat engagements during critical tax weeks in 2025. Agentforce 3 equips leaders to monitor, improve, and scale their AI workforce with confidence.

With a new Command Centre 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.

Why It Matters

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 organisations need to make agent velocity their competitive advantage.

“With Agentforce, we’ve unified agents, data, apps, and metadata to create a digital labour platform, helping thousands of companies realise 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 Centre: Monitor, Measure, and Optimise Your AI Agents 

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 labour. Agentforce Command Centre is that layer: a complete observability solution that gives leaders a unified pane of glass to monitor agent health, measure performance, and optimise 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 optimise your agents. Command Centre empowers teams to analyse 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 Centre 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 centre supervisors can track performance and escalate fast. And over time, every department will have a Command Centre purpose-built for optimising 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 Centre simulates behaviour at scale with state injection and AI-driven evals so you can pressure-test your agents before going live.

“Agentforce autonomously resolved 70% of 1-800Accountant’s administrative chat engagements during the peak this past tax season, an incredible lift during one of our busiest periods. But that early success was just the beginning. We’ve established a strong deployment foundation and weekly are focused on launching new agentic experiences and AI automations through Agentforce’s newest capabilities. With a high level of observability, we can see what’s working, optimise in real time, and scale support with confidence,” said Ryan Teeples, Chief Technology Officer at 1-800Accountant.

Enabling Secure Enterprise Connectivity with MCP and A2A Support

AI agents cannot drive impact if they cannot 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.

Expanding the AgentExchange—a Leading AI Agent Ecosystem—to Make Agent-First Tools Easy to Discover and Use

As organisations embed AI agents across their teams, AgentExchange already makes it easy to equip them with the trusted tools they need to drive value fast with partner-built agent actions and templates. Now, customers will be able to discover MCP servers from 30+ partners that provide trusted access to new third-party tools and resources and seamlessly connect them to Agentforce through a secure AI agent gateway. Launch MCP partners include AWS, Box, Cisco, Google Cloud, IBM, Notion, PayPal, Stripe, Teradata, WRITER and more.

Customers will be able to leverage MCP Servers through AgentExchange from leading partners, including:

• AWS: Agentforce will enable agents to unlock critical insights from vast repositories of unstructured data stored on AWS, leveraging a suite of MCP servers such as the Amazon Bedrock Data Automation MCP Server and Amazon Nova Canvas MCP Server. This will allow Agentforce to intelligently process and summarize documents, extract key information from images, transcribe and analyse audio recordings, identify crucial moments in video files, query enterprise knowledge bases, perform semantic code repository searches, and interact with databases through natural language interfaces via Aurora PostgreSQL/MySQL MCP Servers‌— all enhancing customer service, compliance checks, and operational reporting.
• Box: Agentforce can help retrieve the latest signed NDA from a Box MCP service, including a summary of key terms. Agentforce can then post a Slack update to the relevant teams that includes the AI summarisation, the contract in PDF form, and recommended next steps.
• Google Cloud: Customers can unlock new Agentforce use cases by connecting to key Google and Google Cloud services via MCP. For example, developers will be able to seamlessly connect Agentforce to the Google Maps Platform and generative AI models like Veo, Chirp, Imagen, and Lyria, while also enabling agentic interactions with Google Cloud Databases. This can enable Agenforce actions that span location-based services, creative AI, and transactional data.
• PayPal: Customers can now power a full range of agentic commerce capabilities by connecting Agentforce to PayPal’s MCP server. They can list products, place orders, process payments, dispute claims, track shipping, manage subscriptions, issue refunds, and more with PayPal, making integrations simple and easy.
• Stripe: Agentforce enables AI agents to directly handle a wide range of Stripe functionalities via its MCP server, including real-time product lookups, quick payment link generation, subscription setup, and faster dispute management.
• WRITER: Through a secure connection to WRITER’s MCP server, Agentforce can call WRITER’s enterprise-grade AI agents to automate content generation, knowledge retrieval, and compliance checks all within Salesforce workflows.

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.

Unmatched Industry Readiness Out-of-the-Box Delivers the Fastest Path to Value from AI Agents

Agentforce customers in every industry can deliver value from AI agents fast through more than 200 pre-built industry actions—half of them new this summer—from patient scheduling, to advertising proposal generation, to vehicle servicing, and more. Agentforce 3 also introduces simplified, flexible pricing with new Agentforce for Sales, Service, and Industry Cloud SKUs with per-user pricing and unlimited usage of actions for employee-facing agents to help teams get started fast and scale boldly.

Deploy and Scale Agentforce with Confidence Through Trusted Experts

The Salesforce partner ecosystem, including Accenture, Deloitte Digital, NeuraFlash, PwC, and more, has aided customers across thousands of Agentforce deployments, providing 272,000 certified specialists to deliver strategic guidance and facilitate the successful activation of AI agents that help drive business impact while maintaining governance.

“With Agentforce, PepsiCo is taking a significant step forward in our agentic journey. By unifying our insights through Salesforce Data Cloud, we’re gaining a holistic view of our customers and operations. This visibility enables us to continue to act strategically, strengthen engagement, and drive greater value across every market we serve,” said Athina Kanioura, Chief Strategy and Transformation Officer at PepsiCo.

Available Today

• Agentforce 3
• Agentforce adoption analytics
• Testing Centre 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 authorisation
• 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

Learn More

• See the Agentforce 3 release notes.
• Learn more about Agentforce.
• See the new AgentExchange MCP partner collection.
• Learn about Command Centre.
• Go deeper on Agentforce MCP support.
• See an in-depth product demo about MCP.
• Get started fast with the new Agentforce guide.
• Explore new pricing options for Agentforce.
• Register for Dreamforce 2025.
• Join the Agentblazer Communityto connect with AI enthusiasts and share insights.