Tailor

Adaptive websites

Every visitor sees the page that fits them.

Tailor reads each visitor on every click and puts what they came for first: code for developers, outcomes for buyers, data practices for reviewers. It never hides anything.

See how it works
See use cases

How it works

  1. 1

    Write your versions

    Wrap a section in <Slot> and write one version per audience. Your content, your words.

  2. 2

    Tailor reads the visitor

    Where they came from and what they do, judged by Jev on every meaningful click.

  3. 3

    The page adapts

    Headlines, buttons and section order change off-screen, only when Tailor is confident. Nothing is hidden.

What each visitor sees

Developer
The quickstart, code samples and latency numbers first.
Growth lead
Use cases, outcomes and a walkthrough button first.
Privacy reviewer
What we collect, cookies and retention first.
Investor
Why now, the roadmap and the team first.

We’re in beta and haven’t published lift numbers yet. Built-in holdout measurement is next on the roadmap.

See use cases

Quickstart

# @tailor/react ships in the repo for now (npm package coming soon)
git clone https://github.com/mlndshh/tailor && cd tailor
pnpm install && pnpm dev
<TailorProvider siteKey="site_…" endpoint="/api/tailor/v1/decide" initial={initial}>
  <Slot name="hero">
    <Variant default><Hero /></Variant>
    <Variant audience="developer"><CodeFirstHero /></Variant>
  </Slot>
</TailorProvider>
< 1 s
per decision, shown live in the Lens
≈ $0.00003
model cost per decision (estimated)
0
sections hidden, ever
Read the full docs

Privacy summary

  • Session-only behaviour: clicks and reading time on this site
  • One first-party cookie; no third-party cookies
  • No names, emails or IP-based identity
  • No fingerprinting
  • No inference of protected traits
  • Visitors can see, change and reset every choice
Read our full data practices

Pricing

Tailored for you

Beta

Free

Everything, during the beta.

Tailored for you

Growth

Planned

Usage-based, per 10,000 decisions, with lift reports.

Tailored for you

Enterprise

Planned

Data processing agreement, SSO, regional data handling.

Tailored for you

Unit economics: at about $0.00003 of model cost per decision (estimated), a site making a million decisions a month spends about $30 on inference.

Why now

Per-click decisions were possible with LLMs but too slow and expensive. System One models make adapting every page to every visitor practical. TypeSafe’s published benchmark, per workflow:

LLM workflow

8.566 s

$0.013880

Jev

0.114 s

$0.000081

Read the vision

Questions

Does Tailor hide content?

No. It reorders sections, swaps headlines and buttons, and highlights. Every page and section stays reachable, and the switcher lets anyone pick a view.

How fast is a decision?

Under a second from our server to Jev and back, shown live in the Lens. Changes are applied off-screen, so nothing jumps.

What data do you store?

This session’s clicks and reading time, in your browser’s session storage, plus one first-party cookie holding the current view. Our server keeps nothing between requests.

How do I know it works?

Today: simulated visitors and the Lens. Next: a built-in holdout group that measures lift for each audience.

Why is this possible now?

System One models like Jev return calibrated decisions in milliseconds for fractions of a cent, so deciding on every click finally pays for itself.

Next step for youNew here? See how Tailor works