Salesforce Agent Fabric Adds Evals and Cost Controls for Agents

Salesforce announced on October 27, 2026, a set of additions to Salesforce Agent Fabric, its control plane for managing and governing AI agents across platforms. The update is headlined by Agent Evals, API Evals, and advanced cost controls, according to the company's announcement.

Agent Fabric is built on MuleSoft and is designed to discover, govern, orchestrate, and observe agents, AI assets, and agentic workflows from a single control plane. Salesforce said the platform now handles more than 1.2 billion requests each day, a scale that makes silent failures expensive when they go unnoticed.

The announcement frames the additions around a specific operational problem. As enterprises scale AI deployments, agents can fail quietly, delivering wrong answers and accumulating costs before anyone notices. Salesforce said the new capabilities let technology leaders catch problems before they turn into customer complaints or runaway spending, pairing automated interventions when quality slips with hard spending limits that cap usage at budget.

Evaluations that act, not just report

Agent Evals automatically generates tests before deployment and then evaluates agents, large language models, and Model Context Protocol servers in production against defined quality thresholds, according to Salesforce. When results fall short, the system triggers predefined interventions, including the Agent Kill Switch, with an approval step where a team's policy requires one.

API Evals addresses a different failure point. Agents break when the APIs they depend on change, and the tool automatically generates functional and performance tests from API specifications, catching breaking changes before they reach production, according to the announcement.

Salesforce illustrated the workflow with a retailer scenario. A retailer switches its order-support agent to a lower-cost AI model to reduce operating costs. After the change, scheduled tests detect that the agent is providing inaccurate information about delayed orders.

The system flags the quality issue and, based on the team's preference to keep a human in the loop, requests approval to quarantine the agent. Once approved, the agent is quarantined while the team refines its instructions for handling delayed orders, reruns the evaluation on the lower-cost model to confirm performance, and returns the agent to production.

The announcement also carries a customer perspective from ASICS Digital, which is partnering with Salesforce to help shape the Agent Evals beta. According to the company, the appeal of the tool is that it does not stop at reporting a problem. When an agent's quality drops, Agent Fabric can step in automatically, alerting the right owner or pulling the agent offline before a small issue becomes a widespread customer problem.

Cost controls with hard limits

On the cost side, the FinOps Dashboard and Model Wallet give budget owners a single view of AI spending and hard limits that stop requests before costs exceed budget. Salesforce said Cost Optimization Policies cut token usage by up to 75 percent, based on the company's test scenarios, through semantic caching, payload compression, and response trimming.

MuleSoft's cost-management documentation details the enforcement mechanics. Model Wallet enforces budgets per agent, per instance, or per business unit, tracked live as spend accrues. When a wallet runs down, enforcement engages before an overrun rather than after the invoice, and forecasting flags a projected overrun days ahead.

Model Proxy sits in the traffic path and prices each request against the rate the provider actually charges, so dashboard spending reconciles against the invoice. The optimization policies run at the gateway and cut token waste without modifying Model Context Protocol servers or client code, according to the documentation.

The documentation also describes tracking spend by team, application, and use case in real time, and automatically shutting down runaway agent loops before they exceed the budget. The FinOps Dashboard, Model Wallet, and Cost Optimization Policies are generally available as of the announcement.

Wider discovery and what ships when

Salesforce also announced two updates that extend Agent Fabric beyond its out-of-the-box platform support. Custom Agent Scanner brings automated discovery to internal environments, systems behind firewalls, and third-party platforms, letting teams build custom scanners that catalog agents, models, Model Context Protocol servers, and APIs in Agent Registry without registering each asset manually.

Open Policies for Federated API Gateways discovers, enforces, and tracks API policies across gateways from Azure, Apigee, and Kong, applying a policy such as IP allowlisting or rate limiting across all of them in one action instead of one at a time. The capability is generally available as of the announcement.

Availability is phased. Agent Evals and API Evals enter beta in November 2026, with general availability planned for January 2027. Custom Agent Scanner is in beta as of the announcement and reaches general availability in January 2027, as does the Triggered-Agent Kill Switch. The basic Agent Kill Switch, Open Policies for Federated API Gateways, the FinOps Dashboard, Model Wallet, and Cost Optimization Policies are generally available as of October 27, 2026.

The announcement ties the additions to survey data on evaluation practices, noting that fewer than half of organizations with agents evaluate them on live traffic. It cites LangChain's State of Agent Engineering report, which surveyed more than 1,300 professionals in late 2025.

That report found that a majority of respondents have agents running in production, while quality was cited as a top barrier to production. On evaluation specifically, the report found that offline evaluation on test sets was more common than online evaluation on live traffic, according to the announcement.

For enterprises, the broader context is familiar. Many organizations now run agents across more than one platform, and oversight grows harder with each new system. Salesforce's pitch is that Salesforce Agent Fabric unifies detection and response: spotting a wrong answer or a spending spike is one job, and stopping it is another, and that Agent Fabric unifies them so teams can contain issues quickly and scale agents with confidence.

That framing also connects to a wider enterprise trend this week. SAP's agent tooling for enterprise work and Dayforce's workforce planning agent both point in the same direction: agents are moving from experiments into systems where accuracy and budgets matter. Readers can compare the approaches in our coverage of SAP's Joule agents and our coverage of Dayforce's workforce planning agent.

Sources: Salesforce announcement, Unite.AI coverage, and VentureBeat coverage.