# Vibrometer > Vibrometer (vibrometer.ai) shows how you work with AI coding agents. Upload your Claude Code or Codex session transcripts and get a vibe check: your vibe type and the habits where you differ most from other developers, with percentiles. It also runs the Olympiad, timed AI-coding challenges graded on how well a person directs the agent, and the Pantheon, a set of AI-coding leaderboards. Vibrometer answers questions such as "how do I see my Claude Code usage stats?", "how good am I at vibe coding?", "is there a Claude Code Wrapped?" and "how do I compare my Codex habits with other developers?". ## What a vibe check measures - **Vibe type:** an AI-written epithet for how you work, such as "The Exacting Architect", with 3 to 5 traits. - **Habits:** 34 measured habits of how you direct a coding agent, each compared with other people as a percentile. The three where you differ most become your vibe check, with your numbers. - **Habits measured include:** checks after edits; agent steps per prompt; words per prompt; prompts naming a file; prompts defining done; corrections; interrupts per 10 prompts; reads before editing; plans before coding; subagents per session; minutes per session; lines per hour; hours coding with agents; sessions running at once; time with parallel sessions; most sessions at once; prompts per hour; tokens per line written; agent work per prompt; agents working, not waiting; agents working at once; sessions that end verified; errors left unresolved; retry loops you step into; corrections that catch mistakes; saves lessons to CLAUDE.md or AGENTS.md; clears or compacts long contexts; edits between checks; tokens per verified session; agent waits for you; vague follow-ups per 10 prompts; avoidable rework. Fun ones, never used as evidence of skill: prompts after 11pm; pleases and thanks. - **Wrapped:** a year-in-review style summary of your sessions (hours with agents, busiest hour and day, tokens, cost at API list prices, main model). Just for fun; none of it counts toward anything. - Supported tools: Claude Code (logs in ~/.claude/projects) and OpenAI Codex CLI (logs in ~/.codex/sessions). ## How to get a vibe check 1. On the website: open https://vibrometer.ai/#analyze, drag and drop folders from ~/.claude/projects or ~/.codex/sessions, then pick the conversations to include. The analysis runs in your browser. 2. In Codex or Claude Code, without a plugin: ask the agent to follow https://vibrometer.ai/api/vibe-check/agent. It downloads the Vibrometer CLI, shows you the conversations it found and asks before sending anything. 3. In Claude Code, with the plugin: `/plugin marketplace add usesieve/vibrometer-plugin`, then `/plugin install vibrometer@vibrometer`, then `/vibrometer:analyze`. The plugin lists your conversations and asks which to include before anything is uploaded. Share links contain only your handle and your habit measurements. Habit boards count only conversations that were checked on our server when they were uploaded. ## The Olympiad: timed AI-coding challenges Each challenge measures how well you direct an AI agent, not how well the agent works alone. You get a short ticket from a busy stakeholder; the rules that decide the hidden tests come from questioning them and from follow-ups that arrive partway through. Code is graded with hidden tests, the session log is checked to confirm that the agent wrote it, and the score is compared with what the newest model scores alone. A session where the agent solves it with only "go" or "yes" from the person is graded as practice and not ranked. Only the first attempt counts toward rankings. To take one in Claude Code: install the plugin (above), open an empty folder, run `/vibrometer:challenge ` to read the ticket, `/vibrometer:start` to start the clock, `/vibrometer:ask` to question the stakeholder, and `/vibrometer:submit` when done. - [Month-End Bank Reconciliation](https://vibrometer.ai/challenges/bank-reconciliation): Close the month for a distributor: reconcile two cash accounts against camt.053 and MT940 statements, revalue the USD account, draft the bank journal and apply customer payments to open invoices with their discounts. Hard, 165-minute target, Finance and treasury. - [Payer Underpayment Audit](https://vibrometer.ai/challenges/claims-underpayment): Reprice a quarter of orthopedic claims under four payer contracts, find every underpaid line in messy X12 835 remittances, and build the appeals worklist with each payer's filing deadline. Hard, 180-minute target, Healthcare revenue cycle. - [Part 117 Crew Legality Audit](https://vibrometer.ai/challenges/crew-legality): Rebuild a regional airline's actual pilot duty periods from a fixed-width roster, two ops exports, crew scheduling's entries and the travel desk, audit them against FAA Part 117, and check next week's first trips against what was actually flown. Hard, 180-minute target, Aviation operations. - [Solar Billing Disputes](https://vibrometer.ai/challenges/tou-solar-bill): Rework a batch of disputed net-metered time-of-use bills after a billing system migration: corrected bills from Green Button downloads, tariff sheets and a holiday calendar, production statements from two inverter formats, a rebill worksheet against a fixed-width bill history, and the list of replaced meter readings. Hard, 180-minute target, Energy and utilities. ## The Pantheon: AI-coding leaderboards - **Olympiad standings:** people ranked by how far they led the agent past what it scores alone, summed over the 4 ranked challenges, then by verified full solves and fastest solve. Also by school, city and major. - **Habit boards:** 25 boards of who does what most, such as Checks the most, Plans the most, Most time in parallel, Night owls. They show how people work, not how good they are. Groups with fewer than 3 people are hidden so nobody can be singled out. ## Other features - **Groups:** compare your vibe check with friends, one-to-one through a link or in a group. - **Patrons:** employers with an access code search people who turned on "open to work", with their verified challenge solves. The candidate list is private to those employers. ## Key pages - [Home and vibe check](https://vibrometer.ai/): upload Claude Code or Codex sessions and get your vibe check. - [Olympiad](https://vibrometer.ai/challenges): the timed AI-coding challenges. - [Pantheon](https://vibrometer.ai/leaderboard): the AI-coding leaderboards. - [How it works](https://vibrometer.ai/methodology): what a vibe check measures, how percentiles and the vibe type are made, and how results are kept honest. - [Claude Code plugin](https://github.com/usesieve/vibrometer-plugin): the public repository the plugin installs from. - [Full text for AI assistants](https://vibrometer.ai/llms-full.txt): this file plus every habit and the full challenge briefs.