Agent Evals Workbench

Receipts for coding agents.

Paste or drop a Codex, Claude Code, Cursor, or local-agent trace. Everything runs in your browser: no upload, no backend, no calibrated-benchmark claims.

Support with the $5 Codex run receipt Buy the $39 Browser Operator OS Get a written mini audit Get the full workflow audit

Input

Run trace

local only

Exports