Salesforce has officially launched Agentforce 3, a comprehensive update to its digital labor platform that enables enterprises to manage, monitor, and scale AI agents with enterprise-grade visibility and control. The update follows widespread adoption of Agentforce since its debut in October 2024 and is based on feedback from thousands of real-world deployments across sectors including finance, healthcare, and media.
The new release addresses a core issue that has emerged as AI agent adoption accelerates across industries: a lack of observability. While more organizations are deploying agents to automate repetitive tasks and improve service efficiency, teams often struggle to understand agent performance, trace their decision-making, or make real-time improvements. Agentforce 3 introduces capabilities specifically designed to solve these operational and governance challenges.
“Agentforce 3 is built to manage, measure and optimize AI agents deployed in a new hybrid workforce where humans and agents work side by side,” said Adam Evans, EVP & GM of Salesforce AI.
A Centralized Command Center for AI Observability
A key highlight of the release is the new Agentforce Command Center, a built-in observability layer within Agentforce Studio. The Command Center offers real-time dashboards to track agent adoption, performance metrics, feedback scores, and operational costs. It also includes drill-down analytics for individual interactions and uses OpenTelemetry to integrate with existing observability stacks such as Datadog, Splunk, and Wayfound.
This level of transparency is designed to help companies quickly identify friction points, intervene when agents encounter performance issues, and continuously improve how agents function within complex workflows.
Organizations like 1-800Accountant have already seen tangible benefits. During the 2025 tax season, the company reported that Agentforce autonomously resolved 70% of its administrative chat engagements, significantly reducing support load. With the Command Center, teams were able to monitor agent effectiveness in real time and adjust configurations on a weekly basis.
“We’ve established a strong deployment foundation and are now focused on launching new agentic experiences,” said Ryan Teeples, CTO of 1-800Accountant. “With a high level of observability, we can see what’s working and scale support with confidence.”
Support for MCP Standard and Enterprise Integration
Agentforce 3 also includes native support for the Model Context Protocol (MCP), a technical standard for secure agent-to-application interoperability. The MCP client built into Agentforce allows agents to connect with third-party systems using standard interfaces, removing the need for custom code.
Enterprises can use MuleSoft to convert existing APIs into MCP-ready services, complete with security policies and traffic controls. For teams developing custom MCP servers, Heroku Managed Inference and AppLink offer hosted infrastructure for simplified deployment and management.
The addition of MCP support enables Agentforce to integrate more seamlessly with business-critical systems while maintaining control over identity, data access, and compliance.
“Salesforce’s open ecosystem approach, especially through its native support for open standards like MCP, will be instrumental in helping us scale our use of AI agents with full confidence,” said Mollie Bodensteiner, SVP of Operations at Engine.
Expanding AgentExchange with Prebuilt Industry Use Cases
With Agentforce 3, Salesforce has also expanded the AgentExchange ecosystem, allowing customers to tap into a library of partner-built MCP servers and agent-ready tools. Launch partners include AWS, Google Cloud, PayPal, Stripe, Box, Notion, WRITER, and others—offering plug-and-play access to services like document summarization, payment processing, video analysis, and customer communication.
Some examples of new use cases include:
- AWS: Extracting insights from large unstructured datasets via Amazon Bedrock and Nova Canvas MCP servers.
- Box: Retrieving signed contracts and summarizing key terms directly within Slack.
- Stripe and PayPal: Enabling agents to process payments, issue refunds, manage subscriptions, and handle disputes.
- Google Cloud: Enabling agents to interact with generative AI models (Veo, Chirp, Imagen, Lyria) and services like Google Maps Platform.
In healthcare, organizations such as UChicago Medicine are deploying Agentforce to automate routine patient access center queries, helping staff focus on more complex and sensitive interactions.
Performance Upgrades and Architectural Enhancements
Agentforce 3 is powered by Salesforce’s enhanced Atlas architecture, which brings improved speed, accuracy, and resilience. According to the company, platform latency has dropped by 50% since January 2025. Real-time response streaming is now generally available, enabling faster interactions. Inline citations and web search grounding are also included to support higher-accuracy outputs.
Salesforce has expanded regional coverage to India, Japan, Canada, Brazil, and the U.K., with AI agent traffic now served locally to improve performance and meet data residency requirements. Support for six new languages, French, German, Spanish, Italian, Japanese, and Portuguese—has also been introduced, with more than 30 additional languages in development.
To improve reliability, Agentforce now features automatic model failover, allowing traffic to dynamically shift between LLM providers in the event of latency or service degradation. Support for Anthropic’s Claude Sonnet model, hosted via Amazon Bedrock within the Salesforce trust boundary, is available for industries with stringent compliance needs. Later this year, Agentforce will also support Google’s Gemini models.
For the public sector, Agentforce is now authorized under FedRAMP High and generally available in Government Cloud Plus, allowing government agencies to deploy AI agents that meet the highest federal security standards.
Industry-Specific Actions and Pricing Model
Agentforce 3 launches with more than 200 prebuilt industry actions, including use cases for healthcare (appointment scheduling), manufacturing (service diagnostics), media (proposal generation), and financial services (account updates). These out-of-the-box capabilities are designed to accelerate time-to-value and reduce deployment friction.
Salesforce is also introducing simplified per-user pricing for Agentforce SKUs in Sales, Service, and Industry Cloud. The model allows unlimited usage of employee-facing agents, helping customers experiment and expand without incurring incremental action-based costs.
“By unifying our insights through Salesforce Data Cloud, we’re gaining a holistic view of our customers and operations,” said Athina Kanioura, Chief Strategy and Transformation Officer at PepsiCo. “This visibility enables us to act strategically and drive greater value across every market we serve.”
Salesforce is backing the rollout with a network of implementation partners including Accenture, Deloitte Digital, PwC, and NeuraFlash. More than 272,000 certified professionals are available to help customers design, deploy, and govern agentic solutions in complex enterprise environments.
