Free alpha · sign in to claim your spot

Confidence your AI's changes actually work.

Validity makes your agent verify its own work. It runs the change in your real app and shows what passed, what failed, and what couldn't be verified — before it tells you it's done.

Sign in to claim your spot

Access opens as we add invitees. We're letting people in by hand for now.

coding-agent · ~/app · validity-mcp
Verification reportrun_______
—
How it works

Know what actually works before you merge.

01

Say what "done" means.

Your agent captures the request as a checklist it has to satisfy. Edit it anytime.

02

Build, then verify.

When the agent thinks it's done, Validity checks the change in your real app — not a mockup, your actual UI.

03

Get a verdict.

Pass, fail, or couldn't-verify for each item, each with a one-line reason. No more "looks done."

The Playground · free

See your real components. No backend, no key.

Start with the free Playground. Pin acceptance criteria when you want proof. Gate CI when the team depends on it. Everything runs on your machine.

Your agent drives it
Say "show me my button components," then "now the Login screen dark, mobile" — the canvas switches instantly. Like a design tool, but the source of truth is your code.
Real components, faked world
Your actual providers, auth, and API calls are mocked automatically in an isolated sandbox. Nothing needs a backend to render — including the screens behind login.
Free. No key, no account.
The Playground never phones home and never needs a license. And when you want proof — acceptance criteria pinned to those same renders — verifying against them is free for individuals too. Paid plans are for teams running continuously.
Who it's for

For developers who don't trust "looks good to me" from the agent.

Anyone shipping AI-generated code
A reality check before you merge what your agent claimed it built.
Claude, Cursor, Codex users
Your agent verifies its own work before handing it back, so 'done' actually means done.
Free during alpha

“Done” should mean actually done.

Validity is in a free, invite-only alpha. Sign in to claim your spot. We're letting people in by hand and will send the install command when yours opens.

Free during the beta. The continuous watcher becomes part of a paid plan later — one-off verification and reports stay free.