How AI agents
show their work.
A small, curated collection of real AI agent interfaces, each one dated, sourced, and paired with a single design insight you can apply to your own product.
Collection 01 · October 2026
Three ways agents ask, warn, and recover.
The first collection comes from the moments where an agent touches the outside world. It asks permission to run a tool, asks before it opens a URL, and explains itself when a firewall stops it. Each one is a real screen from official documentation, with one idea worth borrowing.
Allow once is the big button; wider trust sits one click deeper
A split button can hold a whole permission ladder while making only the narrowest rung the reflex click.
Split-button approval
Two approvals for one fetch: going out, and coming back in
Split consent where risk changes direction: contacting a site risks data going out; using its reply risks instructions coming in.
Split request/response approval
When the agent is blocked, it says what, why, and how to unblock it
Report a blocked action as a receipt (what was blocked, which command, the two narrowest ways forward) so a security control reads as a next step, not a failure.
Blocked-action explanation
The index
The questions this collection answers.
Concept · illustrative, not a real product
Deep dive: how a pattern could evolve
A composite example for teaching, kept apart from the documented collection.