Salesforce introduces Agentforce 3, an updated version of its platform for digital assistants, including an observation and management module. The update aims to help companies deploy AI agents in a scalable and controlled manner.
Salesforce expands its AI offerings with the launch of Agentforce 3. The platform is designed to help companies manage AI agents more efficiently and deploy them more widely. Central to this release is the new Agentforce Command Center, which provides insight into the behavior, performance, and impact of AI agents in the workplace.
Control over AI activities
The Agentforce Command Center enables tracking and adjusting the entire journey of AI agents. Companies can use dashboards to see how agents are performing, what mistakes they’re making, and where improvement is possible. The Command Center integrates with existing tools via OpenTelemetry, and is compatible with Datadog and other monitoring solutions.
Agentforce 3 also supports the open Model Context Protocol (MCP), allowing companies to easily connect existing tools with AI agents without custom code. Salesforce enables creating, managing, and securing MCP servers via MuleSoft and Heroku. This allows AI agents to perform actions within trusted corporate environments.
Extensive support
With more than 100 pre-built industry actions and support for six new languages, Agentforce 3 aims to deliver value faster in sectors such as healthcare, finance, and retail. Customers like PepsiCo, Grupo Globo, and UChicago Medicine are already using the solution for customer interaction, internal support, and process automation.
Through the revamped AgentExchange platform, customers gain access to a network of more than 30 technology partners, including AWS, Google Cloud, Box, Stripe, and IBM. These offer ready-to-use MCP servers for various applications, such as document analysis, e-commerce, or content creation.
read also
Salesforce Adds dx to Agentforce 2: Proactive AI and Low-Code Tools
The underlying Atlas platform also received an update. The infrastructure now supports response streaming, automatic failover between AI models, and lower latency, among other features. Salesforce also offers support for using models such as Claude from Anthropic and Gemini from Google within its own trust domain.
Agentforce 3 is available immediately, along with new pricing options with unlimited use for employee-focused agents. Additional features such as session tracking and broader adoption analytics will follow in the summer.