find where your bill leaks.

One-week, fixed-price snapshot of your AI coding bill. You hand over read access to one repo and 30 days of agent logs. You get back a benchmark report, three concrete wins, and an honest call on whether deeper work is worth your money.

• Benchmark report. Your codebase run through bench/ with three representative tasks (refactor, feature add, bug fix). Tokens per task, tool calls per task, dollar cost per task. Ranked by waste. Reproducible — fixture committed to your repo so you can re-run after fixes.
• Top 3 wins. Three concrete code-graph queries that would have replaced the worst retrieval failures we found. Before/after measured on the same tasks. If we can't find three wins, we say so — and the audit ends with the memo, not the pitch.
• Pro-fit memo. Honest call on whether a full Pro engagement is worth it for you. Two pages, no slides. If it's not a fit, we say so and tell you what would unblock it. You can show the memo to your CFO without us in the room.

1. Day 0 — email us

   Send the repo URL (or upload a tarball), your agent of choice (Cursor, Claude Code, Codex CLI, OpenCode, etc.), and 30 days of agent usage logs if you have them. We respond within one business day with a Stripe link.

2. Days 1–4 — we run the benchmark

   We build a code graph against your repo (read-only, in our environment), pick three representative tasks from your recent agent history, and run them through both your current retrieval setup and a graph-first setup. Tokens, tool calls, and dollars get measured both ways.

3. Day 5 — you get the artifacts

   Benchmark report (markdown + raw numbers), top-3 wins (with PR-ready code-graph queries), and the pro-fit memo. Delivered as a private repo you keep. No NDA required on our side; we'll sign yours.

• Not a sales call dressed as an audit. The pro-fit memo will say "don't hire us" if it's true. We'd rather lose a Pro lead than a referral.
• Not bespoke integration work. We don't write production code into your repo during the audit. The wins are demonstrated on a fork or in a sandbox.
• Not for stacks with no real AI coding usage yet. If your team is still evaluating whether to use coding agents at all, the audit will land flat. Talk to us when you have a month of agent history.
• Not refundable after Day 1. Once we have your repo and start running the benchmark, the work is done. If you're not sure it's worth $1,500, the open-source version of Coograph plus the public benchmark on the home page will tell you most of what the audit would.

We don't gate the request behind a form. Send an email with the fields below and we reply within one business day with a Stripe link. If we can't take you on this month, we tell you when we can.

Include: company / team name, repo language(s), team size, approximate monthly AI coding spend, and which agents you use day-to-day. That's enough for us to write back with a useful next step. If you'd rather skip the audit and go straight to a Pro engagement, see /pro/.