Salesforce Expands Agent Fabric With Agent Evals And API Evals
Salesforce expanded Agent Fabric on October 27, 2026, adding Agent Evals, API Evals, advanced cost controls, custom agent discovery, and cross-vendor API governance. The changes target organizations operating AI agents across Salesforce, internal systems, third-party platforms, and multiple API gateways. The Agent Fabric announcement separates features available now from capabilities scheduled for beta and general release.
Agent Evals connect testing with intervention
Agent Evals will generate tests before deployment and evaluate agents, large language models, and Model Context Protocol servers in production against defined quality thresholds. When results fall below policy requirements, Agent Fabric can trigger predefined interventions, including alerts, approval workflows, or an Agent Kill Switch.
Agent Evals enters beta in November 2026, with general availability planned for January 2027. The evaluation-triggered kill switch is also scheduled for general availability in January. The existing Agent Kill Switch is generally available now.
API Evals test dependencies before they break agents
API Evals generates functional and performance tests from API specifications. This is intended to detect breaking changes in the APIs used by agents before those changes reach production.
API Evals follows the same rollout schedule as Agent Evals, with beta planned for November 2026 and general availability planned for January 2027. Teams using agents that depend on frequently changing internal or external APIs will need to treat API specifications as part of the agent testing process once the capability becomes available.
Cost controls and discovery features are available sooner
The FinOps Dashboard, Model Wallet, and Cost Optimization Policies are generally available now. Model Wallet provides budget controls for AI usage, while the dashboard consolidates spending visibility. Cost Optimization Policies apply techniques such as semantic caching, payload compression, and response trimming.
The Custom Agent Scanner is available in beta immediately. It extends Agent Fabric discovery to internal environments, systems behind firewalls, and third-party platforms, allowing teams to catalog agents, models, MCP servers, and APIs in the Agent Registry without manually registering every asset.
Open Policies for Federated API Gateways is also generally available. It allows organizations to discover, enforce, and track policies across Azure, Apigee, and Kong gateways from a centralized control plane. Policies such as IP allowlisting and rate limiting can be applied across supported gateways instead of configured separately on each one.
