Salesforce has announced a major expansion of Agent Fabric, a control plane designed to manage and govern a diverse landscape of AI agents and large language models. By establishing a centralized layer for orchestration, the system allows businesses to optimize model selection based on cost and performance rather than routing every task to a single expensive resource.
Andrew Comstock, SVP and GM of MuleSoft, stated, “Agent Fabric gives you total oversight over your AI landscape. It allows you to turn a fragmented collection of agents into a unified, high-performance digital workforce.” This expansion also brings the technology to new regions, including Japan and Canada, supporting both private cloud and on-premises workloads. The updated platform introduces guided determinism through Agent Script, which enables organizations to define fixed handoff rules while allowing AI to handle the underlying reasoning.
A new visual authoring canvas simplifies how developers map workflows and human checkpoints, while the AI Gateway standardizes token management and cost controls across multiple vendors. To enhance security, the system now features trusted agent identities that require mobile approval for high-stakes operations like financial transactions. John Pettifor, SVP of Innovation at Diabsolut, noted, “Agent Fabric is how we scale AI without losing control.” By integrating these governance tools, the firm ensures that every agent interaction remains auditable and aligned with vetted business rules.