AI coding agents are now everyday tools. They read code, run commands and write changes all day. Every one of those steps uses tokens, and tokens cost money. Until now, most teams could not answer a simple question: how much are our agents actually using, and where?
Today DarkControl adds token usage tracking. You can now see your agents' token consumption in the same console where you already see what they did and which policy decisions applied.
What's new
- Token usage per agent activity. For agents that report usage, DarkControl records the tokens used and keeps them with the rest of the activity record.
- One view for everyone. Usage sits next to the allow / ask / deny decisions and the audit trail, so you don't need a separate tool or spreadsheet.
- Breakdowns that answer real questions. See usage by user, team, agent and model, so you can see where consumption comes from.
- Estimated cost where the agent reports it. Finance can follow spend trends without asking engineering for screenshots.
For engineers: find the waste
Efficiency starts with visibility. With token usage next to each session, you can spot the patterns that burn budget quietly:
- Sessions that run far longer than the task needed.
- Agents stuck in retry loops, trying the same blocked action again and again.
- Heavy models used for simple jobs that a lighter model could handle.
- Teams or repositories whose usage jumps suddenly, which is worth a quick look.
Because usage and policy decisions live in one place, you can also see the link between them. An agent that keeps hitting a deny rule wastes tokens on every attempt. Adjusting the rule or the instructions saves money as well as noise.
For finance: predictable AI spend
AI agent costs are often spread across many developer accounts and vendor invoices. That makes them hard to forecast and hard to allocate. Token usage tracking gives finance a clear, regular picture:
- See usage trends over time instead of waiting for the monthly invoice.
- Attribute usage to teams and departments for internal reporting.
- Spot unusual spikes early, before they become budget surprises.
- Base planning conversations on recorded data, not estimates.
Why this belongs in a governance platform
Governance is not only about blocking risky actions. It is about knowing what your AI agents do, on whose behalf, and at what cost. Frameworks such as ISO 27001 and SOC 2 expect organisations to monitor the tools they depend on. Usage data adds the resource side to the security and compliance record you already keep with DarkControl.
It also keeps things simple. You don't need another dashboard, another vendor or another data pipeline. The same lightweight endpoint agent that applies your policies now helps you understand your consumption too.
Getting started
Token usage tracking is available in your DarkControl console. Coverage depends on what each AI coding tool reports, and the console shows clearly which agents provide usage data, so a missing figure is never mistaken for zero.
New to DarkControl? Start with a free 7-day watch-only audit on up to 10 devices. Nothing is blocked and no credit card is needed. Or book a short demo and we'll show you token usage alongside your agents' activity.
