Revenium has introduced Guardrails, a new set of runtime controls that enables enterprises to enforce AI spending limits and model access policies at the moment an AI request is made. Complementing Revenium’s Tool Registry and AI Outcomes capabilities, Guardrails makes real-time decisions about whether AI calls should be allowed, helping organizations govern AI usage, control costs, and enforce security policies before resources are consumed.
Visibility and outcome tracking tell a team what already happened. Neither one reaches into the moment a developer’s script calls an unapproved model, or a support workflow quietly drifts into a pricier tier. By the time a dashboard shows the spike, the money is already spent. Revenium’s Guardrails moves that decision earlier by checking a rule against a call before it reaches the provider, instead of after it lands on a bill.
“Teams don’t want to wait for a budget review to decide whether a brand-new AI model belongs in their stack,” said Jason Cumberland, CPO and co-founder, Revenium. “With Guardrails, that decision is a rule instead of a policy nobody reads. Point it at a model like Claude Fable 5, set it to enforce, and the answer is already built into the workflow.”
With Guardrails, teams can scope a rule to a specific organization, product, agent, model, or task type, then decide how it responds. A rule can alert a team through the channel it already watches, or block the call outright before it reaches the provider.
Rules can also start from a filtered view of spend by employee, carrying that same scope over automatically, and every rule keeps a history that teammates can check without the ability to make changes. When a call gets blocked, whoever set the rule can attach a message so a developer sees an explanation instead of a dead end.The model-level scope gives teams a clean way to handle a new release.
When Anthropic shipped Claude Fable 5 this summer as the newest model in its top-tier Mythos class, many engineering leads wanted their teams to stay on the existing lineup until the new tier had a reviewed price tag and an approved use case. A Guardrail scoped to a single model closes that gap without a code change. Point one at Claude Fable 5, set it to enforce instead of notify, and every call that agent or application makes through Revenium’s SDK stops before it reaches the provider until someone decides otherwise.
“Every dashboard we’ve built has been about knowing what already happened,” said John Rowell, CEO and co-founder, Revenium. “Guardrails is the first one that decides what happens next. A rule can stop a call before the provider ever sees it, which is a different kind of control than a chart that updates the next morning.”
The release also adds smarter cost-risk alerts, watching for spend risk before the bill arrives by flagging when cost per call climbs faster than usage, often a sign a model swap or prompt change quietly makes every call pricier, and by catching when a single agent, product, or API key breaks from its own normal spending pattern. When a day’s AI spend jumps,
Revenium now explains why on its own, tying the spike to the specific people and usage behind it and weighing that against what the team actually produced that day. Every number across the product is now labeled as either a real provider invoice or Revenium’s own observed usage, with a reconciliation view showing where the two line up and where instrumentation may be falling behind actual spend. Employee-level usage can also be filtered by model tier, provider, and vendor, benchmarked against team norms and cost per million tokens, and exported to CSV.
Learn more at https://www.revenium.ai/
Related News:
