GetWrangler.ai Dashboard →

GetWrangler / Cost Audit

See what you're overpaying before you change anything.

Paste a usage sample or upload a CSV. No signup, no routing changes — just the number. Nothing you submit is stored.

Drop a billing export CSV exported from OpenAI or Anthropic's usage dashboard, or model/tokens/cost rows you paste yourself

If this includes usage from Claude Code, Codex, or similar coding tools: Codex usage in OpenAI exports is detected and excluded automatically. For Claude Code usage billed through Anthropic, mark the relevant API key(s) below before running the audit — GetWrangler doesn't route or discount that kind of traffic, and leaving it in (Anthropic only) will distort your real savings estimate.

Nothing you paste is stored. This runs once and the result is discarded.
One CSV with a header row (model, tokens or input_tokens/output_tokens, cost) — or, for Anthropic, both the Cost and Tokens exports together. See the "Export from your console" tab for exact steps.
Files are parsed in your browser. Only the parsed rows are sent, never the file itself, and nothing is stored.

OpenAI

  1. Go to platform.openai.com/usage.
  2. Click Export, then select the "Activity data" tab — not "Cost data": cost data has no per-model breakdown and isn't needed for the required file.
  3. Set API capability to cover your actual usage (e.g. "Responses and Chat Completions").
  4. Under "Group by", make sure Model and API Key are both checked — the default "All selected" is fine and simplest.
  5. Set the date range to the last 30 days, Time interval: 1 day, File format: CSV.
  6. Click Download, then use the Upload file tab above to upload it.

Optional, for a rough total-spend sanity check only: also export the "Cost data" tab the same way (same date range) and upload it alongside the Activity file. Model isn't groupable on that tab regardless of settings, so it can't feed the per-model savings math either way.

Anthropic

  1. Go to the Claude Console at platform.claude.com/cost (Anthropic's billing dashboard). Set the range to the current month (or last 30 days), group by Model, and click the download icon to save the cost CSV.
  2. Go to platform.claude.com/usage, use the same date range, and click the download icon to save the tokens CSV.
  3. Anthropic's export splits cost and token data into two separate files — that's expected, not a mistake.
  4. Upload both files together on the Upload file tab above (select both at once, or drop them in one after another).
  5. We'll match them up by date and model automatically. If a few rows can't be matched, we'll say so rather than silently dropping them.
Uses the same upload path as any other CSV — nothing about this is treated specially, and nothing is stored.
Want this running continuously? Track every call, verify savings live, and control exactly what gets routed where.
Start free →

This audit estimates savings from cost-appropriate model routing based on the sample provided. It does not modify your existing API calls or infrastructure. The public classifier distinguishes fewer categories than GetWrangler's full account-level pipeline, so the estimate here is a lower-fidelity preview, not a guarantee. Full continuous tracking, verification, and savings-share billing apply only once you connect an active account.