AI Workflow Score

AI Workflow Score

See how you really work with AI.

One command. One honest number. A breakdown that shows what to improve.

npx @beon-tech/ai-score

About a minute · your code never leaves your machine · your score goes on the board

ReadsClaude CodeCodexCursorGitHub Copilot
Your readinglast 30 days · 192 sessions
80OUT OF 100
Leverage
25/30
Craft
16/20
Verification
12/15
Delivery
12/15
Consistency
7/10
Customization
8/10
  • ✓ Checks ran in 80% of sessions
  • Next: ship more of your sessions
dimensions, one score
6
point scale
100
command to breakdown
~1 min
lines of code sent
0

Challenge a friend

Think they can beat your score?

Send your link. They see your number and the breakdown, get the command, and find out where they land next to you on the board.

Checking your session…

What it measures

Not how much you use AI.
How well.

  • 30 pts

    Leverage

    Whole tasks delegated, not autocomplete.

  • 20 pts

    Craft

    Clear prompts. Failures carried to a fix.

  • 15 pts

    Verification

    Tests, typechecks, builds actually run.

  • 15 pts

    Delivery

    Work lands as a commit or a PR.

  • 10 pts

    Consistency

    Steady weekdays, not a last-week burst.

  • 10 pts

    Customization

    MCP, skills, hooks, subagents.

How it works

Three steps, about a minute.

  1. $ npx @beon-tech/ai-score login
    Opening ai-score.beon.tech/device…
    Your code: WXYZ-4821
    Waiting for approval ▂▃▅
    Step 1

    Sign in from your terminal

  2. ai-score.beon.tech/device

    Approve this device?

    WXYZ-4821

    Approve
    Step 2

    Approve the code, run the scan

  3. Your readingprivate link

    80/ 100

    Next: more sessions ending in a commit.

    Step 3

    Read your breakdown

Privacy

We count what happened. We never read what you wrote.

Your machine

stays here, always

sends

AI Workflow Score

counts and yes/no states only

  • your promptshow manywords per prompt
  • your code & diffslines added / removed
  • commands & their outputchecked? yes/nodelivered? yes/no
  • the agent's repliestool callstokens per modelsession timing
  • file names & pathsproject folder, hashedwhere agents keep logs
  • repo names & branchescommits madePRs opened

Your name, score and city go on the public board. Nothing else about you gets published. The CLI is open source, so you can check every line of this. Read the full privacy policy and terms.

Who's behind this

Built by BEON.tech, in Latin America, since 2018.

BEON.tech is a software talent company founded in Buenos Aires in 2018. It places engineers from across Latin America with U.S. companies and has screened more than 68,000 of them along the way. This score grew out of the question its team asks every day: not whether someone uses AI, but how well.

About BEON.tech

  • 2018

    founded in Buenos Aires

  • 68,000+

    engineers screened across Latin America

  • 23

    Latin American countries engineers are hired from

  • Claude Partner

    Select Services Partner in Anthropic's partner network

  • Where's the leaderboard?

    At /leaderboard: every verified engineer's latest score, with country and city filters.

  • What's a good score?

    Bands are calibrated on real engineers' runs. The breakdown tells you what to work on first.

  • Can I run it again?

    As often as you like, on the machine you actually work on.

Ready to see your score?

npx @beon-tech/ai-score

About a minute. See where you land on the board.