One brand colour in. A palette out where every pairing is measured.

PairPass is one joined workflow, not four separate tools: generate a role-assigned palette that passes by construction, audit every pairing by the job it does, repair the failures without changing your hue, and export tokens that contain only the pairings that pass — free, with no account and nothing uploaded.

1. Generate by construction

   Your colour becomes primary and neutral scales in OKLCH, with roles — background, text, action, border, focus — solved against the colours they will sit on, in light and dark themes at once.

2. Audit by use

   The matrix grades every pairing for its actual job: body text at 4.5:1, large text and components at 3:1, focus rings against the surface they cross — not one ratio for everything.

3. Repair, hue held

   A failing colour moves in lightness only until the re-measured ratio clears its threshold. The hue — your brand — does not move, and the whole palette re-grades after every fix.

4. Export only passes

   CSS variables, Tailwind v3 and v4, and design-token JSON whose semantic layer points only at passing pairings. Failures are excluded and named in the file, never shipped silently.

The workflow

Export doors

Why another contrast tool?

Because the pieces exist separately everywhere and joined almost nowhere. Generators produce colours first and check them afterwards — if at all. Checkers grade one pair at a time and leave you to guess which colour to change. Exports happily ship the failing pairings alongside the passing ones. And the number behind it all is routinely rounded in your favour: a ratio of 4.496167 displays as "4.50" in most tools and quietly passes, although the standard is explicit that 4.5 means 4.5. PairPass is built around the opposite habits, and its methodology page publishes the formula and the hand-computed test record so you can check the arithmetic yourself.

The problem is not rare: the WebAIM Million 2026 audit found low-contrast text on 83.9% of the top million home pages — the most common detected accessibility failure, again (WebAIM, webaim.org/projects/million, February 2026 data, checked November 2026).

Guides