Know which team, customer, and feature drove your AI bill

The invoice arrived. Now finance wants answers.

Your OpenAI or Anthropic bill jumped 40% this month. Finance asks the obvious question: which team, which customer, which feature caused it? You open the provider's dashboard — and it shows you one number: the total. There's no per-team line, no per-customer line, no way to tell whether it was the new feature you shipped or one power-user hammering an endpoint.

So the spike becomes a manual investigation: grep logs, guess at traffic, email around. By the time you have an answer, it's next month and the question has changed.

TOLVYN attributes every request — so the bill explains itself

TOLVYN sits as a proxy in front of your providers. Every request that passes through is metered and tagged, so spend is attributed the moment it happens. Instead of one total, the dashboard breaks the bill down across the dimensions that actually matter:

Each of these has its own breakdown view in the dashboard, plus a filterable request log. Requests can also carry X-Tolvyn-Feature and X-Tolvyn-Agent tags, which are recorded on every request and used for attribution and filtering.

How it works

• Proxy. Point your traffic at TOLVYN — one base-URL change, or a drop-in SDK for Python, Node, or Go. It forwards to OpenAI, Anthropic, or Google unchanged.
• Tag. Add X-Tolvyn-* headers (or set them once in the SDK) to mark each request with the team, service, user, or customer it belongs to.
• Meter. TOLVYN counts tokens using the provider's own reported usage and computes cost in microdollars with exact decimal arithmetic — not floating-point estimates.
• Attribute. The cost is recorded against the request's tags, metadata only — prompts and responses are never stored.
• View. The dashboard shows spend broken down by team, service, end customer, user, and model, with a request log you can filter and export to CSV.

One honest caveat

Attribution is only as good as your tags. A request that arrives untagged is still metered accurately, but it rolls up to the organization total as unassigned — it won't appear under a specific team or customer until you tag it. Tagging is a one-time setup per service, and the SDK makes it a single line.

What this is — and what it isn't

TOLVYN gives you the attributed cost data: the accurate, per-dimension breakdown of what your AI actually cost and who drove it. That's the input finance needs for cost allocation, COGS-per-customer, and chargeback conversations.

TOLVYN is not a billing system — it does not issue invoices, charge your customers, or replace your provider's billing. It answers "who spent what, and can I prove it?" — and because every record is written to a tamper-evident audit ledger, the numbers hold up when someone asks you to back them.

Stop guessing where the bill went

Tag your traffic once, and the next invoice explains itself. Free to start — 10,000 requests a month, no card required.