CoreStack Launches TokenOps to Govern AI Economics
By: Ezhilarasan (Ez) Natarajan, Founder and CEO of CoreStack
Today, CoreStack is launching TokenOps as an expansion of our FinOps+ offering to govern AI cost and usage alongside cloud. As tokens become the new unit of enterprise spend and agents take on increasingly important work across business applications, TokenOps lets enterprises discover the providers, models, and agents across their AI estate, control which models and services their teams use, attribute spend to appropriate business units, and prove which investments deliver value. Organizations can buy TokenOps on its own or add it to FinOps+ to govern AI alongside their multi-cloud estate under the same attribution model, policy library, and evidence trail.
In most enterprises, AI governance is divided, with 41% of executives saying ownership varies by department.1 Typically, FinOps teams track what AI costs and security teams track what it touches. TokenOps is built to see the full picture, allowing teams to govern both from a single platform.
CoreStack introduced TokenOps in October as a module within FinOps+ to give teams visibility into AI spend across Amazon Bedrock and Anthropic. This release expands TokenOps from visibility and cost attribution to unified AI governance.
Governance Has to Keep Pace with AI Advancement
AI is moving faster than the controls around it. New models arrive every month, and autonomous agents now carry out tasks across business applications with less human oversight. Organizations need visibility into AI use across users and teams, control over access and permissions, and clear accountability for the actions agents take.
McKinsey’s State of AI Report found that 62% of organizations are experimenting with agents, and 23% are already scaling them in at least one business function.2 Once agents are in production, governance must extend to which models a team can call, which agents are authorized to act, and what data can enter a prompt to ensure AI use remains within approved guardrails. The pace of innovation means that controls have not kept up, with 77% of organizations saying AI adoption is outpacing their governance capabilities.3
“Most enterprises now run AI across dozens of providers, assistants, and agents, yet few leaders can say what’s in use, what it costs, or whether it’s operating within policy. Visibility was the first step. Governance that matters is enforced before a request executes, not discovered after the bill arrives,” shares Ez Natarajan, Founder and CEO at CoreStack. “TokenOps brings AI under the same governance model enterprises use for their cloud estate, as part of a single platform for governing cloud, SaaS, and AI.”
What’s New in TokenOps
TokenOps now has its own dedicated experience and dashboards to allow for AI-specific visibility and attribution. Deep discovery now includes Microsoft Foundry, including Foundry Agents, in addition to Amazon Bedrock and Anthropic. Enterprises can reconcile spend to their Azure bill, allowing teams to apply chargeback, budgets, and anomaly detection just as they do for cloud. This release also expands service analytics, including cost, prompt, unit economics, and optimization insights to more than a dozen AI tools employees use every day, including OpenAI, Claude Code, and Microsoft Copilot Studio.
New guardrails enforce policy at the point of consumption. Model allow-lists and access permissions determine which models a team can call before a request executes, to keep usage and spend within approved models instead of catching violations after the bill arrives. TokenOps also tightens controls on sensitive data, with PII, toxicity, and bias controls applied through AWS and Azure gateways that enterprises already route traffic through, ensuring sensitive data is blocked before it reaches a model without deploying additional infrastructure.
Prompt and agent analytics, built on OpenTelemetry, show which prompts and agent runs are driving cost. They also give security and compliance teams an auditable record of how AI is used across the organization. Access is controlled by the same role-based access control (RBAC) model used across the CoreStack platform, with only authorized users with the right roles being able to view prompt-level data.
This release also includes AI unit economics, which measure cost per user, session, pull request, or accepted line of code, so FinOps and Finance leaders can tell which AI investments are paying off, retire the ones that aren’t, and defend continued funding to the board with defensible data. Since AI governance is an ongoing program rather than a one-time project, the new AI Governance Maturity Scorecard gives leaders a single baseline across attribution, unit economics, optimization, budgets, and anomaly detection. With the scorecard, they can measure and report governance progress over time with confidence.
“Knowing what you spend on AI services is the easy part. The hard part is knowing what an AI workload really costs, which agent and which user drove it, whether it was worth it, and whether it stayed inside policy,” says Parthu Kishen, Senior Vice President of Product at CoreStack. “An AI workload is more than model tokens. It runs on cloud infrastructure, SaaS and other services too. TokenOps resolves that full cost down to each agent and each user behind the workload, and enforces guardrails like model allow-lists and PII detection at the gateway. That is what makes an AI investment defensible to a CFO and the board.”
Request a demo to see how CoreStack TokenOps can help your organization govern AI cost and usage from a single platform.
