Salesforce has unveiled Agentforce 3, a significant upgrade to its digital labour platform designed to help enterprises scale their use of AI agents with greater visibility and control. The announcement builds on insights from thousands of deployments since the platform’s launch in October 2024.
The latest release introduces new capabilities such as the Agentforce Command Center for observability, native support for open standards like Model Context Protocol (MCP), and an expanded AgentExchange marketplace. These enhancements are intended to address enterprise demands for monitoring tools, secure integrations, and improved interoperability as the adoption of AI agents grows.
Improving observability through the Agentforce Command Center
A key feature of Agentforce 3 is the Command Center, which offers real-time observability across all AI agent activities. It allows business leaders to monitor agent health, track performance metrics, and intervene quickly when necessary. Dashboards provide insights into agent adoption, success rates, and cost efficiencies. This data is captured through an extensible session-tracing model built into Salesforce’s Data Cloud, using the OpenTelemetry standard to integrate with platforms like Datadog, Splunk and Wayfound.
Built into Agentforce Studio, the Command Center also includes development tools powered by natural language. Teams can quickly generate test cases, simulate agent behaviour at scale, and assess performance through AI-driven evaluations before launching new capabilities.
“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,” said Ryan Teeples, Chief Technology Officer at 1-800Accountant. “With a high level of observability, we can see what’s working, optimise in real time, and scale support with confidence.”
Expanding secure connectivity and partner access
Agentforce 3 strengthens support for interoperability with the inclusion of a native MCP client, enabling out-of-the-box connectivity to any MCP-compliant server. This allows enterprises to give their AI agents secure, controlled access to internal tools and third-party services without requiring custom code.
Salesforce has also expanded the AgentExchange marketplace, which now includes more than 30 partners such as AWS, Box, Cisco, Google Cloud, IBM, Notion, PayPal, Stripe, Teradata, and WRITER. Organisations can easily discover and connect to MCP servers for tasks like document summarisation, customer support automation, e-commerce fulfilment, and database queries.
Use cases include AWS’s suite of MCP servers for unstructured data processing, Google Cloud integrations with generative AI models, and PayPal capabilities such as order processing and subscription management. With WRITER, Agentforce users can automate enterprise-grade content generation and compliance checks directly within Salesforce workflows.
Enhancing platform performance and enterprise readiness
The upgraded Atlas architecture underpins Agentforce 3, delivering lower latency, higher accuracy, and improved reliability. Salesforce reports that system latency has dropped by 50% since January 2025. Real-time response streaming is now generally available.
Customers can also choose from a wider range of hosted large language models (LLMs), including Anthropic’s Claude Sonnet model via Amazon Bedrock, hosted within Salesforce’s trust boundary. Support for Google’s Gemini will follow later in the year.
Salesforce has also introduced over 100 new pre-built industry actions this summer, expanding the total available to more than 200. These span use cases from healthcare scheduling to automotive servicing. The platform now offers simplified per-user pricing and unlimited action usage for internal-facing agents across sales, service, and industry cloud products.
“With Agentforce, PepsiCo is taking a significant step forward in our agentic journey,” said Athina Kanioura, Chief Strategy and Transformation Officer at PepsiCo. “By unifying our insights through Salesforce Data Cloud, we’re gaining a holistic view of our customers and operations.”
The platform’s growth is supported by Salesforce’s partner ecosystem, including Deloitte Digital, Accenture, PwC, and NeuraFlash, which collectively offer over 270,000 certified professionals to assist with deployment and integration.
“Over the past several months we’ve listened deeply to our customers and continued our rapid pace of technology innovation,” said Adam Evans, EVP and GM of Salesforce AI. “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.”
