A free plugin for Claude Code · v0.6.0

Your agent says "done".proof makes
sure it
is.

Coding agents move fast. proof double-checks their work before it reaches you: pages load, buttons do what they should, screens match the design. If something's off, your agent fixes it first.

See how it works
The problem

Fast work still needs checking.
Usually by you.

A button that doesn't respond. A page that never loads. A screen that's almost, but not quite, the design. Small things slip through, and you're the one who finds them.

The button doesn't respond

It looked finished. Nobody clicked it.

The page never loads

Fine in the code. Blank in the browser.

The screen is almost the design

Wrong word, a few pixels off, the wrong grey.

How it works

Your agent checks its own work.
Before you ever see it.

Each time your agent finishes, proof checks the app still works. If something's off, the agent sees exactly what and fixes it before handing back.

Fails go back to your agent

Your agent gets the error and keeps working. You see the reply once the checks pass.

Commits are gated

Before every git commit the full checks run. A failure blocks the commit.

Checks can't be loosened

If your agent edits proof.json, proof stops it and asks you first.

What it catches

Seven ways "done" can slip.
proof catches all seven.

Flip the switch to see each one without proof, then with it.

what-it-catches.test7 passed · 0 failed

Your agent says "done"

…and the tests fail when you run them.

Checks run after every reply that changes files. Your agent sees the failure and fixes it before handing back.

A failing check gets switched off

…so the work "passes".

Editing proof.json, settings.json or the fidelity kit makes your agent stop and ask you first.

The screen is almost the design

Wrong copy, a few pixels off, the wrong grey.

Every tagged element is checked against the mockup: exact copy, position and size within 2px, font, colour.

The empty state never got built

The spec didn't mention it.

Every state needs its own mockup frame, and every requirement says how it'll be verified.

A test run hangs

Your agent waits forever.

Test runs and builds have time limits. A breach is reported as a hang.

Your laptop freezes

Three agents and a simulator at once.

Heavy runs wait for memory. If the machine goes over the ceiling, the newest run is stopped.

You come back after an hour

No idea what happened.

The dashboard shows every session: what's running, for how long, the plan and the check results.

From idea to finished

From idea to working app.
Checked at every step.

Your agent plans it, designs it, then builds it. Each step has to pass before the next one starts, and a second agent double-checks the work before it's saved.

01

Discover

Idea → spec

/proof:discover
Every requirement says how it's verified
02

Design

Spec → HTML mockup

/proof:design
Every screen, state and test ID exists
03

Build

Any code change

/proof:build
Tests come first, the app is run, screens match the mockup
The dashboard

Walk away for an hour.
See what happened at a glance.

/proof:watch shows every session live: what your agents are working on, how far they've got, and whether everything still works.

The proof dashboard: a build with three agents running, its tasks, features, gate checks and memory, and where the time went

Needs you, in coral

A session that stopped on a question stands out from the ones still working.

Every agent in one run view

Lanes, verifiers and explorers, with timers and commits.

Local only

Listens on 127.0.0.1. proof makes no network calls.

Several sessions at once

Run three agents at once.
Your laptop keeps up.

proof keeps an eye on memory and makes the heavy jobs take turns, so your computer never freezes.

Memory, one machinesample
68%16.3 of 24 GB
Your appsYour agents' runsNext run
1

Heavy runs and new agents wait

until there's room.

2

If the machine goes over anyway

the newest run your agent started is stopped.

3

Your apps are never closed

proof tells you which ones are using the most.

Let your agent build.
proof makes sure it works.

Paste one prompt into Claude Code. It checks for git and Node.js, asks before installing anything, and sets proof up for you.

Read the docs on GitHub