Script-tag signup forms

Try it live → Run the signup-form demo with your own API key.

The shape

You have a long signup form. Users bounce. You want to replace it with a friendly chat that asks one question at a time, validates as it goes, and posts the collected data to your existing endpoint.

Linda’s script-tag mode does exactly that, in 10 seconds.

Markup

<script src="https://unpkg.com/@linda/script/dist/linda.js"></script>

<form
  id="signup"
  data-linda
  data-linda-config="/configs/signup.json"
  data-linda-provider="anthropic"
  data-linda-model="claude-sonnet-4-6"
  data-linda-proxy="/api/linda"
  action="/api/save"
  method="POST"
>
  <input name="email" required />
  <input name="company" />
  <input name="role" />
  <input name="team_size" type="number" />
  <input name="goal" />
  <button type="submit">Submit</button>
</form>

That’s it. Linda hides the form, mounts a Shadow-DOM chat overlay, and asks the user one question at a time. When all fields are collected, it un-hides the form (or auto-submits, your call).

Config

{
  "persona": "Friendly onboarding assistant. One question at a time. Warm, brief, never robotic.",
  "target": { "type": "form", "selector": "#signup" },
  "dataPoints": [
    { "name": "email", "question": "What's your work email?", "validate": "email" },
    { "name": "company", "question": "Which company are you with?" },
    { "name": "role", "question": "What's your role there?" },
    { "name": "team_size", "question": "Roughly how big is your team?", "validate": "number" },
    { "name": "goal", "question": "What are you trying to do with us?" }
  ],
  "completion": { "type": "submit-form" }
}

What this gets you

• One question at a time. No 12-field wall of inputs.
• Real validation. Linda’s beforeFieldFill hook can veto or transform.
• Existing handlers run. Your form’s submit listener fires normally on completion.
• No backend changes. Your /api/save endpoint sees the same shape as if a human typed.
• Streaming UX. Token-by-token response, native ReadableStream.

Why not just use a chatbot widget?

A widget would ask the same questions but throw away the answers when the user closes the tab. Linda writes to your real <input> fields, so your form’s existing validation, React state, browser autofill, and Save-the-form-state behavior all work normally. The agent is operating on the form — not replacing it.

What it costs

Per session: ~5–10 LLM round trips × ~500 input tokens × Sonnet 4.6 ($3/M input). $0.0075 to $0.015 per completed signup, including prompt caching savings. Way cheaper than the human time saved.

FAQ