// ONE LOCAL TRANSCRIPT · ZERO UPLOAD

What happened during this run?

We report observed human text turns, tool calls, errors and the longest uninterrupted run. These are activity measurements, not a quality score.

→

Run it on yours

The CLI can also run a current-state check. The browser reads only the transcript, so its outcome is always marked not checked.

terminal
npx isitanagent --verify --cwd . --card -- npm test

Runs the exact executable and arguments after -- without a shell. A zero exit code means only that this check passes in the current workspace now. Attribution remains not verified.

npx isitanagent --card

Skip the check for an activity-only card marked OUTCOME NOT CHECKED.

no terminal
Select a .jsonl session log drop it here or click to pick one · it stays in this tab

Usually in ~/.claude/projects/. In the file picker, press Cmd+Shift+G and paste that path. The browser parses it locally.

There is no upload endpoint in this site to upload to. The link the tool prints carries numbers in the URL fragment, which browsers never transmit to any server. Turn off your wifi and score a file — it still works.

Inspect the privacy and execution boundary Don't trust it. Check it.
npx isitanagent --dry-run

Prints the exact blast radius, then exits without reading a byte:

  • READS ~/.claude/projects/*/*.jsonl — and names the single file it would score
  • WRITES nothing by default. --card writes one SVG in your current folder.
  • NETWORK nothing. No fetch, no telemetry, zero dependencies.

npm view isitanagent dependencies returns {}; the package has no install scripts; and the default and --card paths work with wifi off. The only network paths are the explicit --contribute and --forget commands.

A verifier is user-selected executable code. It may read files or use the network even though isitanagent itself opens no socket.

One run is not an experiment

The same model and prompt can produce a different run. A useful comparison needs repeated trials from the same clean baseline, one changed factor, the same protected verifier and a fixed budget. Compare pass rates first. Treat activity counts as secondary.

Read the comparison protocol →