Stacklet launches Token Custodian to govern AI token spend

Stacklet has launched Token Custodian, a control plane for governing AI token usage. The product is entering early preview as companies face growing pressure to account for AI spending.

Token Custodian attributes token consumption across teams, agents and projects, then applies policy and workflow rules to manage spend in real time. It is aimed at FinOps and platform teams that want to monitor usage across multiple AI providers and tie costs to specific teams, applications and cost centres.

The launch comes as businesses look for ways to show a return on rising AI budgets. Stacklet cited research from the Tokenomics Foundation's State of Tokenomics 2026 survey showing that 43% of respondents identified proving value as their biggest challenge, while a much smaller share pointed to price.

Under Stacklet's approach, organisations can set budgets and usage rules by team, project or environment. When usage approaches a limit, the system can respond by shifting workloads to a lower-cost model or sending an approval request, rather than simply blocking activity.

Cloud heritage

The product builds on expertise Stacklet developed through Cloud Custodian, an open-source cloud governance project under the Cloud Native Computing Foundation. The company says the same policy-based automation used in cloud governance can now be applied to AI token usage.

Kevin Emamy, VP of Development for the Tokenomics Foundation, linked the launch to a broader need for stronger financial controls around AI.

"Organisations are moving quickly to scale AI, but the ability to govern that investment and connect it to business value is still catching up," he said. "As a Tokenomics Foundation member, Stacklet brings expertise in cloud governance to this broader effort to help enterprises manage AI economics and turn AI investment into measurable business outcomes."

Analysts say the issue is not only the size of AI bills, but whether companies can distinguish between spending that supports useful work and spending that does not. That has become more relevant as enterprises experiment with AI agents and coding tools that can generate substantial token usage.

"Applying the same cost limit to AI agents that deliver significant business value and those that consume tokens without comparable returns can prevent organizations from realizing the full potential of agentic AI," said Torsten Volk, Principal Analyst, Application Modernization at Omdia. "Token Custodian enables FinOps teams to define and enforce more granular controls for token usage, helping enterprises direct spend toward workloads where it makes business sense and manage costs where it doesn't. This closes the critical gap between simply reporting on AI spend and managing it based on business impact and priorities."

Early customer view

One early user named by Stacklet is Avalara. The tax software company's Cloud and AI Optimization team is evaluating the product as it looks for tools that fit existing working practices.

"Governing our AI spend so it drives value, not waste, is exactly what we're after, and most tools either stop at reporting or weren't built for FinOps and governance workflows. Token Custodian does both, and it fits right into the tools our teams already use, like Claude Code and Slack. We're excited to try it," said Lindbergh Matillano, Director of Cloud & AI Optimization at Avalara.

For Stacklet, the product extends a governance model from cloud infrastructure into AI operations. Its customers already manage more than USD $10 billion in combined cloud and AI spend through its software, and the company sees token governance as a related challenge for enterprise technology teams.

Travis Stanfield, Chief Executive Officer of Stacklet, said the focus should be on directing spending rather than simply reducing it.

"The instinct with AI spend is to cut or cap it, but that just slows your teams down," he said. "The real win is governing it so spend flows to the work that pays off, and that's what we built Token Custodian to do, on the same engine enterprises have trusted to govern their cloud for years."