WorkExperimentApr 2026
Agent Ledger (wireframe)
A static wireframe of an operations console for supervising AI agents: workflow queue, credit spend, outputs and a command channel.
Agent Ledger: left rail of agents, a workflow queue with credit use in the centre, and an inspector and command channel on the right
- Agent Ledger collapsed into stacked sections on a narrow screen
01What I built
An afternoon wireframe from 2026-04-18 of what a calm, local operations console for an agent workforce might look like: a left rail with agents and link status, a main workspace with the workflow queue, current credit use, final results and an activity map, and a right rail with the selected workflow's details and a command channel. Selecting a workflow updates the inspector; typing a command appends to the channel. Warm off-white surfaces, hairlines instead of cards, status colours that are functional rather than decorative.
02What I learned
This is the sketch that became the Mission Control section of the ELO Agent Operating System proposal two months later. Two design questions I wrote down at the time are still open: should it feel like a system monitor or an agent inbox, and should credits be budget guardrails or cost estimates. Wireframing it first is why the later proposal timeboxed the real build to two days.
03Built with AI
Built in a live preview loop: describe a change in chat, refresh the browser, repeat. No framework, so the whole thing is three files.