How to Run a Claude Team Spend Audit (Free, in Your Browser)

If your Claude bill has quietly grown and you’re not sure where the money’s going, a spend audit is the fastest way to find out — and you can do it in about a minute, for free, without handing your data to anyone. Here’s exactly how.

Audit your team's Claude spend, free →

What you’ll need

• Owner (or Primary-Owner) access on a Claude Team plan.
• Your Anthropic Spend Report — the per-user, per-model CSV of usage and spend.

That’s it. No install, no signup for the headline result.

Step 1 — Export your Anthropic Spend Report

In claude.ai:

1. Go to Settings → Analytics.
2. Click Export Spend Report.
3. Pick a window — Last 90 Days is a good default.

You’ll get a CSV with one row per person × model: usage volume (requests and tokens) and spend. (Note: this is the per-user Spend Report from claude.ai — not the API/Console billing export, which doesn’t have per-person detail. If you grab the wrong one, the audit will tell you and point you to the right export.)

Step 2 — Run the 8 checks (in your browser)

Head to the free audit and drop the CSV in:

→ Run the free spend audit

The whole thing runs on your device — your spend file (which carries employee emails and dollars) is never uploaded or stored. It parses locally and runs 8 checks:

1. Dead-weight seats — over-tiered and idle seats you could right-size or reclaim.
2. Spend concentration — how much of the bill a few people drive (the “your top user costs 7× the median” reveal).
3. Model mix — how much is going to expensive models (e.g. Opus running where a cheaper model would do).
4. Hidden metered surfaces — line items admins often don’t realize are billed.
5. $/request outliers — the sessions that cost far more than the team median (often runaway agents or huge context).
6. Discount / list exposure — how much of your spend is masked by credits (renewal-relevant).
7. Allocation by domain — employees vs contractors at a glance.
8. Context ratio — a coarse proxy for prompt-heavy waste.

You’ll see an instant headline number first (no email), then the full per-person breakdown.

Step 3 — Act on it (you’re in control)

Everything the audit shows is a recommendation you confirm — it never changes anything in your account. The usual moves:

• Right-size over-tiered seats (Premium → Standard) and reclaim dormant ones. (To turn “uses Standard-level volume” into a downgrade, pair it with your seat roster — the audit is explicit about that. More on seat tiers and how to spot the overpay.)
• Nudge model mix — move cheap tasks off the expensive model.
• Investigate the $/request outliers — that’s where runaway spend hides.

Step 4 — Put it on a cadence

Team composition drifts every month, so a one-time cleanup drifts back. The audit takes a minute — re-run it monthly (you can opt in to a figure-free reminder) and the savings stick.

A couple of honest caveats

• On seat-based plans, within-allotment usage is the flat seat fee, so some dollar figures are overage-only and every projection is a labeled estimate (“should closely match your invoice”). The audit leans on request/token volume, which is always present.
• Exact Premium→Standard right-sizing needs your seat roster (the CSV gives usage, the roster gives who’s on which tier).
• It’s a snapshot, not monitoring. Catching a spike as it happens is a different (paid, coming-soon) job; the free audit is the point-in-time read.

That’s the whole process. It’s independent (we don’t sell Claude seats), free, and your data never leaves your browser.

→ Run the free spend audit — or see WTClaude for Business, and the broader July 2026 pricing picture this audit fits into.

FAQ

How do I audit my team’s Claude Code spending? Export the Anthropic Spend Report CSV (Settings → Analytics, as an Owner), then run it through the free browser-based audit — 8 checks, nothing uploaded.

Where do I get the Anthropic Spend Report? claude.ai → Settings → Analytics → Export Spend Report; pick your window. Requires Owner/Primary-Owner on Team.

Is it safe to upload my Claude spend report? With WTClaude’s audit nothing is uploaded — it runs entirely in your browser, so the file never leaves your device.

What does the audit actually check? Eight things: dead-weight seats, spend concentration, model mix, hidden metered surfaces, $/request outliers, discount exposure, domain allocation, and context ratio.

---

WTClaude is a free, open-source, billing-grade Claude Code cost tracker — independent, not affiliated with Anthropic. Run the spend audit — 8 checks, in your browser, nothing uploaded.