Salesforce adds hard AI budget limits and a generally available kill switch to Agent Fabric
A FinOps Dashboard, Model Wallet and Cost Optimization Policies are generally available now, along with the Agent Kill Switch. Agent Evals, which tests agents in production and can trigger the kill switch, enters beta in November and reaches general availability in January 2027.
Salesforce announced new cost controls and evaluation tools for Agent Fabric, its control plane for managing and governing AI agents across platforms, on November 1. The company says Agent Fabric handles "over 1.2 billion requests" daily. Some of the new pieces are generally available now, while the evaluation tools arrive in beta this month. Salesforce did not give pricing in the announcement.
What is available now
The cost features are generally available today. Salesforce says "The FinOps Dashboard and Model Wallet give budget owners one view of AI spending," with "hard limits that stop requests before costs exceed budget." Cost Optimization Policies, it says, "cut token usage by up to 75% (based on test scenarios)" through semantic caching, payload compression and response trimming. That figure is Salesforce's own, from test scenarios rather than customer deployments.
The Agent Kill Switch is also generally available, as are Open Policies for Federated API Gateways, which apply a policy such as IP allowlisting or rate limiting across Azure, Apigee and Kong gateways in one action. A Custom Agent Scanner, which discovers agents, models, MCP servers and APIs behind firewalls and on third-party platforms and lists them in the Agent Registry, is in beta now.
What comes later
Agent Evals generates tests before deployment and then checks agents, LLMs and MCP servers in production against quality thresholds. When results fall short, it triggers predefined interventions, including the Agent Kill Switch, with approval where a team's policy requires it. Agent Evals and API Evals, which builds tests from API specifications, are in beta in November 2026 and generally available in January 2027. The Triggered-Agent Kill Switch also arrives in January 2027. Salesforce's own example is a retailer that moved an order-support agent to a cheaper model, where evaluations caught inaccurate delayed-order answers before the agent was returned to production.
"Enterprises can only scale agents as fast as they can keep them accurate and on budget," said Andrew Comstock, SVP and GM of MuleSoft and Agent Fabric. The announcement carries Salesforce's standard caution that customers should base purchase decisions on features that are already released.
Why it matters
Hard spending limits at the request level are a different control from a monthly invoice review: they stop work when the money runs out. That protects budgets, and it can also stop a customer-facing agent mid-task. Before enabling them, decide which budgets map to which agents and who is alerted when a cap is reached. The automated link between failed evaluations and the kill switch is the more ambitious part, and it is not generally available until January.