Your CFO is going to ask.
Most engineering leaders don't have the answer.
CostHawk shows you what the dev team shipped with AI, what it cost, and what drifted somewhere else.
Every dev team's AI spend is going up.
Most technical leaders can't tell you which of it is tied to the roadmap and which has drifted.
CostHawk can.
costhawk.ai
Uber burned through their entire 2026 AI budget in four months.
Their board is not happy. Half the engineering teams in this country are one quarter away from the same problem.
CostHawk is the answer. Real usage. Real output. Receipts your board can't argue with.
CostHawk.ai
Codex usage is up 180% MoM. Claude Code is still leading, but the gap is closing.
Live from the CostHawk leaderboard.
Which tool is your team on?
costhawk.ai
A CFO's favorite question in 2026:
"What did we get for the $400K we spent on AI last quarter?"
A CTO's least favorite question in 2026:
"What did we get for the $400K we spent on AI last quarter?"
The 3-step CTO survival guide for 2026:
Step 1: Buy AI tools for the whole dev team.
Step 2: Don't track where the tokens are going.
Step 3: Hope your CFO never asks.
How to lose your CTO job in 3 easy steps:
1: Pay for AI tools across your dev team.
2: Don't measure what the tokens build.
3: Wait for your CFO to ask where the money went.
I built @CostHawkAI to track what AI agents actually cost. One command in Claude Code: "Show me my full usage and spend breakdown."
64,275 API calls. $249 across 3 providers. 6.3B tokens through Claude Code.
Retail equivalent? $12,960. I pay $200/mo for Claude Max.
Most devs have no idea where their tokens are going. One command to find out:
npm exec --yes costhawk@latest -- --login
Free at costhawk.ai
Your AI spend is a black box. You know the total. You don't know the breakdown. That's how you end up overpaying for months without realizing it. Open it.
👀 Codex is rapidly climbing against Claude Code, accelerating 34x faster this month.
Come join our free anonymous leaderboard. Track your CLI usage, get real-time pricing, and see how you stack up.
costhawk.ai
I've been building @CostHawkAI and I'm happy to keep it free. One thing the public leaderboard is showing pretty clearly right now: Codex is accelerating 17.1x faster than Claude Code this month. 1,266.8% MoM vs 74.0%. Codex 5.4 is very good.
costhawk.ai/leaderboard