Tool call inspector
Every tool call behind an agent's answer, with request and response, and a claim check that marks each sentence as supported, unsupported or inference. Point at a sentence and a line runs to the exact field that proves it; verify a made-up figure and the sentence corrects itself.
Agent workspacesNo. 2032 variationsFree with an account
Playing on its own
Variation
- The idea
- Every tool call behind an answer, with each claim checked against the data
- Autoplay
- Four tool calls land in the inspector while the answer fades in; Check claims marks each sentence as supported, unsupported or inference; hovering the delay sentence draws a line to the max_delay_days field of its call; the unsupported revenue figure opens a popover, Verify runs a new billing call, and the sentence strikes $18k and writes $14.2k; the loop ends on another sourced sentence.
- When you click
- Toggle Check claims (or press C); hover, click or Tab to sentences to see their source; Verify (V or Enter) an unsupported sentence; expand any call to read its request and response, or rerun it. On a phone the inspector is a bottom sheet that opens on the call behind the sentence you tap.
- Look
- chat plus inspector, 10 px cards, hairlines, drawn link line, FLIP cards, quiet fades, sans with mono JSON type
- Variations
- Carbon, Paper
- Files
- index.html, style.css, script.js and prompt.md. No framework, no CDN, works offline.
The prompt and the files
Paste the prompt into Claude, ChatGPT, Lovable or any coding agent to rebuild it in your own product, or take the three files as they are.
Loading the files