vibe check v1.0.0
Jan 1 to Jan 13 (12 days) · OTel traces (1 session)
Your vibe, level 3 of 5

Agent wrangler

57/100

You know your way around the agent, and it mostly knows its way around you. A couple of leaks still cost real money every week. Patch them and you're a level up.

Level 1 · 0 to 19Prompt rouletteLevel 2 · 20 to 39Token bonfireLevel 3 · 40 to 59Agent wranglerLevel 4 · 60 to 79Claude whispererLevel 5 · 80 to 100AI superfactory

Your radar

8 checks, each scored from 0 to 100. 100 is the best score, so a bigger shape means better vibes.

Vibe check radar: one score per check, 0 to 100, where 100 is the best scoreNo idle rebuilds0Lean context58Lean smalledits91Short agent loops97Clean tool runs52Right modelfor the job92Clearprompting34Lean setup32
  • No idle rebuilds0$85.38
  • Lean setup32$21.54
  • Clear prompting34$39.11
  • Clean tool runs52$4.92
  • Lean context58$56.29
  • Lean small edits91$6.67
  • Right model for the job92$2.38
  • Short agent loops97$22.72

Your potential

1.8× the work on the same plan≈$219 a month, 46% of your usage
Improvement 1 · idle cache rebuilds

Start fresh after a break

The exact fix for your setup is in vibe check Pro.

≈$131a month, 27% of your usage, at API prices, estimate
Improvement 2 · context bloat

Keep context lean

The exact fix for your setup is in vibe check Pro.

≈$78a month, 16% of your usage, at API prices, estimate
Improvement 3 · tool errors

Stop tool retries at the first failure

The exact fix for your setup is in vibe check Pro.

≈$8.31a month, 2% of your usage, at API prices, estimate

Based on $193 traced at API list price · 320M tokens · 976 calls · 1 session. No bill to compare with, so coverage is unknown. Prices: API list price, checked 2026-10-09. Cache lifetime: 1 hour (inferred: the cache survived 7 of 7 pauses of 5 to 60 min). Every dollar figure is an estimate.

Get the Pro version coming soon

This free report reads what is on this machine, or only this one session in the cloud. Pro is planned to go further.

  • Your whole history, or any timeframe you pick, across all your sessions, local and cloud, instead of what is on this machine or in this one session.
  • Every check in detail and your habits, with the exact fix for your setup.
  • Your sessions and models, broken down.
  • Also available for your entire developer team.
Get notified

Every check in detail

Available in Pro

Your habits

Available in Pro

Sessions and models

Available in Pro
How we calculate

Every number comes from code, never from a language model. Costs are tokens times the published per-model prices, taken from one price file with a checked date. Savings are simulated one after another on a copy of your call log, so they never double count. The vibe score is the average of the check scores (0 to 100, 100 is best).

Not modelled: batch discounts, the 1.1× US-only data residency surcharge, and Amazon Bedrock and Google Cloud prices.