HN Showcase
Back to browse
GitHub Repository

Pre-flight cost estimation for Claude Code. Know what your AI task will cost before it runs.

15 starsTypeScript

Tarmac – Know what Claude Code will cost before you run it

by sarthakaggarwal·Mar 5, 2026·2 points·1 comment
Visit ProjectView on HN

AI Analysis

●●●BangerWizardrySolve My ProblemBig Brain

Conformal prediction trained on 3K tasks hits 81% cost accuracy.

Strengths
  • •Addresses genuine pain: $2 estimate turning into $15 bill
  • •81% in-range accuracy on 80% target—rigorous validation, not hand-wave
  • •Zero setup friction: hook install, local inference, no external API calls
Weaknesses
  • •Useless for non-Claude-Code workflows (Agentic Cursor, other IDEs excluded)
  • •Depends on SWE-bench representativeness; real-world task diversity may vary
Target Audience

Claude Code users managing AI coding budgets

Similar To

CloudTrail cost calculators · Token counting tools

Post Description

I built this because I kept getting surprised by Claude Code bills. A complex refactoring task that I expected to cost $2 would end up costing $15, and there was no way to know beforehand.

Tarmac analyzes your prompt and gives you a cost range before Claude Code starts working. It uses conformal prediction trained on 3,000 real SWE-bench tasks — hitting 81% in-range accuracy on an 80% target.

npm install -g tarmac-cost — fully open source, runs locally, no tracking, no accounts.

Would love feedback on estimate accuracy for your use cases - that's what I'm most focused on improving

Similar Projects