Save money on Claude Code with 41 vibe check.
vibe check is a free skill for Claude Code that shows where your tokens and money go, and the 3 areas worth improving. It runs in Claude Code on your computer.
people know their vibe alreadyFree skill for Claude Code. Analysis runs locally.
41 does this for the AI agents your company runs. Become an early adopter.
Eight checks, each from 0 to 100. 100 is the best score.
Could save 46% of usage, about $219 a month at API prices. On a plan: 1.8× the work.
How it works
From download to savings, inside Claude Code.
Download and install
Step 1: download the ZIP, or clone it from GitHub (you get the link after sign-up). Step 2: install the skill: tell Claude Code to install it, or unzip it into ~/.claude/skills. Step 3: run /vibecheck. Under a minute.
Click here to get the skill.
See your level
Type /vibecheck. You get your level and how much you could save: a share of your usage limit, or dollars on the API.
Improve and save
The report shows your radar, what you could get back and the three areas worth improving, with what each could save.
Type /vibecheck, see your level and top 3 areas to improve. The free skill runs on your machine. It reads only and sends nothing.
Five levels. Which one are you?
Your vibe score puts you on one of five levels. 100 is the best score.
0 to 19 pointsLevel 1: Prompt roulette
Sometimes it works, mostly it doesn't, and you pay for every spin.
20 to 39 pointsLevel 2: Token bonfire
Things get built, and a lot of tokens go up in smoke on the way.
40 to 59 pointsLevel 3: Agent wrangler
You know your way around the agent, and it mostly knows its way around you.
60 to 79 pointsLevel 4: Claude whisperer
Lean sessions, warm caches, few retries.
80 to 100 pointsLevel 5: AI superfactory
Short sessions, the right model for each job, almost nothing wasted.
Free, and Pro for teams
Start free. vibe check Pro is coming soon: one plan for one developer or a whole team.
vibe check Free
$0Runs on your machine
- Your level and vibe score
- The radar of eight checks
- The top 3 areas to improve, with savings, and the badge to share
- Your sessions on this computer, last 30 days by default. Added to a cloud session: only that session
- Runs local only, nothing leaves your machine
Free for every developer. Installs in under a minute.
vibe check Pro
for developer teams$5 per developer / month+ $100 / month base fee per developer team
- Everything in Free
- Not just a skill: works with your Claude Code telemetry data (tokens, costs, models)
- Full history and any timeframe, across machines, cloud sessions and your team
- Every check in detail, your habits, and the fix for your spend
- Your whole team, with a works-council mode
$10 $5 early-adopter price per developer
The Pro dashboard
Coming soon. What a team sees: spend over time, every developer's level, the areas worth improving.
Spend over time
Per week, with per model and per project views.
Top fixes
- Start fresh sessions sooner$640 per month
- Use a smaller model for simple edits$410 per month
- Stop re-reading the same large files$260 per month
Developers
| Developer | Level | Score | Could save |
|---|---|---|---|
| Darth Debugger | Level 4 | 81 | $95 |
| Samwise Gitgee | Level 5 | 92 | $40 |
| Captain Pushard | Level 3 | 64 | $210 |
| Princess Lintia | Level 2 | 47 | $390 |
| Doctor Whomerge | Level 3 | 58 | $275 |
Works-council mode: on. Every developer shows under a playful nickname, so no real person can be identified in the dashboard. Switched off, managers see each developer by name. It is designed for works-council agreements.
Privacy (& what changes with Pro)
The free skill reads only and sends nothing. Pro is different, and we say how.
Free skill
- Runs on your machine. It has no network code, and tests fail the build if any appears.
- Numbers, never words: with the free skill, what you wrote becomes counts and is never stored or sent.
- Every number comes from code, with published prices and a "last checked" date.
Pro (coming soon)
- Pro is different: usage counts will leave your machine and will be stored in 41's cloud.
- Pro will be hosted in the EU on AWS. Other regions on request.
- Raw data will be kept 13 months, then only daily summaries.
- Everything will be deleted within 30 days after cancelling.
- Managers will see each developer, or nicknames in works-council mode.
FAQ
What does the skill show?
Your level and vibe score, a radar of eight checks, what you could get back (the share of your usage limit on a Claude plan, dollars per month on the API), the top 3 areas to improve, with savings, and the badge to share. Team reports are part of Pro.
How much history does it read?
Every Claude Code session on your computer that Claude Code still keeps: by default the last 30 days. vibe check Pro will keep your full history and any timeframe you pick.
Which setups does it work with?
Claude Code on your computer, with any login: claude.ai, an API key or Bedrock. No claude.ai account is needed. The skill only reads the Claude Code session files on your disk.
Does it work in Claude Code cloud sessions?
Yes, but only for the session you are in. A cloud session holds just its own history, so the report covers that one session and nothing before it. To see more, run it in Claude Code on your computer, where it reads the last 30 days by default. In Pro, cloud sessions will work with more history.
I'm on a Claude subscription, not the API. Does it work?
Yes. On a plan, vibe check shows how much of your usage limit you'd get back and how much more work the same plan does. Dollar figures are shown at API prices for reference.
Does any of my data leave my machine?
With the free skill, 41 never receives anything from it. The skill reads your Claude Code session files where Claude Code runs and has no network code. Its summary appears in your Claude conversation like any other answer. The only thing 41 gets is the email address you enter here to receive the download link.
What happens with my email?
You get one mail with the download link and the install steps, and after sign-up you also see a GitHub link to clone the skill instead. Every mail has an opt-out link. See the privacy policy.
How are the numbers calculated?
In code, never by a language model: tokens times published prices, from one price file with a "last checked" date. Savings are simulated on your own sessions. Every dollar figure is an estimate.
What is vibe check Pro?
Coming soon. vibe check Pro adds your full history and any timeframe you pick, every check in detail, your habits, the exact fixes for your setup, your sessions and models broken down, and a team view for your whole developer team. There is one plan for one developer or a whole team: $100 per month flat plus $5 per developer per month, an early-adopter price (the regular price is $10 per developer). Join the Pro waitlist to hear when it opens, or find out more at get41.ai.
What leaves my machine with Pro?
Pro is not open yet; this is how it is designed. With Pro, usage counts will leave your machine: tokens, costs, models, timings and tool names. Claude Code's built-in OpenTelemetry sends them to 41's collector, and they will be stored in 41's cloud (EU, on AWS). You turn it on with one settings snippet per developer, or one managed setting for the whole team. Only counts are sent. Prompts, answers, code and file names will be discarded on arrival, if they are ever sent, so they are never stored. The free skill still sends nothing.
What is the works-council mode?
A Pro setting that will hide who is who. Each developer will show under a playful nickname from geeky films and series, so managers see usage without seeing a real person. It is designed for works-council agreements.