The best way to make quick adjustments
CmdZero maps every element in your running app to its exact line of source. Click something, change it — copy and style edits land in your code instantly, no AI involved. Everything else goes to Claude Code, on a Claude model sized for the job.
The 40 second tweak
You know the loop. You ask your agent for "a bit more padding on the pricing card." It greps the repo, reads four files, thinks, and 137,000 tokens later edits the right one — probably. You look at it and say "hmm, a bit less."
CmdZero uses AI at the right time.
How to get started
$ npm i -D cmdzero
@cmdzero/react// tsconfig.json { "jsxImportSource": "@cmdzero/react" } // app/layout.tsx <CmdZeroOverlay />
$ npx cmdzero # open your app, press ⌘0 # and click anything
No bundler config. Works with Turbopack, webpack, and Vite — stamping happens in React's dev JSX runtime, so server components map too. Production builds get a passthrough: zero bytes, zero overhead.
Automated tool selection
Click, type, done
Click any text and edit it in place. Enter writes it to the JSX literal — your actual source file, not a CMS shadow copy. 0 tokens, 0 API calls.
Your design system, on tap
Padding and margin per side. Font sizes and swatches read from your Tailwind theme — the tokens your app really uses, not a generic picker. Every change is a deterministic class edit.
A direct path to context
Every element in you UI is source mapped to provide minimal sufficient context for the model to understand the request.
The right model for the task
CmdZero already knows which file you clicked, so nothing has to search for it — and it routes each tweak to the cheapest model that can still do the job.
The file is already known. No grep, no repo crawl — just the component you clicked and your instruction.
Simple tweak, cheap model. Every request is routed to the smallest model that can still handle it well.
Automated AI cost savings
The cheapest request is the one that never hits a model. CmdZero skips the API when it can, right-sizes when it can't, and shows you exactly where the savings add up.
Some tweaks cost nothing. Cached and deterministic edits resolve locally — no tokens, no API call, no bill.
Small tasks, small models. Flagship models stay reserved for hard work — routine edits never pay premium prices.
See where you save most. Alerts surface which files, models, and tweaks are driving the biggest cost cuts.
The ultimate refinement tool
Select across containers, reorder sections by hand, and watch every change reload live — while a running tally shows exactly what you're saving.
Grab many at once. Select across containers and apply one tweak to every element you picked.
Reorder by hand. Drag sections into a new order and the layout rearranges as you go.
See it instantly. Every tweak reloads the page automatically — no manual refresh, no lost state.
A running tally. Watch cumulative savings climb as skipped and right-sized requests add up over the day.
Know the moment it happens. Real-time alerts stream in for spend spikes, model routing, and cache hits.
Your next tweak takes four milliseconds
Press ⌘0 and click anything.