Turn complex forms into executable workflows.
Onboarding, supplier checks, claims: many people and AI agents, many steps, weeks apart. Write it once, in React. Clipless runs it: it sends each person their link, chases them when they're late, shows them only what they may see, and records every answer and decision as an event.
- React
- Next.js
- TanStack Start
- React Router
- Astro
- Vite
Any React 19 app. Your components, your models, your backend.
A submission isn't a database update. It's an event.
A form builder saves one person's latest answers and hands you a spreadsheet. The approvals, send-backs and checks after that end up in email and Zapier, and what someone told you in March is overwritten in July. Clipless runs the part after submit and keeps all of it.
Where Clipless has four stars, there's more to build.
One person, once? A form builder is the right tool. Clipless is for the forms that pass between people and AI agents.
Define it once. Everyone gets their own form.
The whole process lives in one React file, so a change is one edit in one place. Each person gets a form with only their stage in it, on a hosted page they're emailed a link to or inside your app. The AI step's prompt stays on the engine, out of the browser.
<Workflow id="expense_claim_v1"> <Stage name="claim" actor="role:employee"> <Field name="description" required /> <Field name="amount" type="currency" required> <Remarks /> </Field> <Action id="policy" type="ai_prompt" blocking prompt="Is {{description}} within policy?" schema={VERDICT} /> </Stage> <Stage name="review" actor="role:manager" renderIf={{ 'policy.flagged': { eq: true } }} transitions={[SEND_BACK]}> <Field name="decision" type="select" options={DECISIONS} required /> </Stage> <Stage name="payout" actor="role:finance"> <Field name="payment_ref" required /> </Stage> </Workflow>
-
claim
Employee
fills it in
-
policy
AI agent
checks the policy
-
review
Manager
decides
-
payout
Finance
pays it out
The AI takes a step. A person signs it off.
An AI agent is a participant, not a plugin. It reads the document, fills in what it can, flags what's off and suggests the rest for a person to accept. Its answers land on the same record as everyone else's, so six months later you can see what it was shown, what it said and who accepted it. Fields you mark as personal data never reach the model.
Mar 4 09:12 employee claim amount $182, receipt dinner.pdf Mar 4 09:12 ✦ AI policy flagged: over the $75 meal cap Mar 4 11:40 manager review decision send back, remark "split the bill?" Mar 6 11:40 engine claim reminder sent: 2 business days Mar 6 13:05 employee claim round 2: amount $74 Mar 6 13:05 ✦ AI policy within policy Mar 6 13:05 ✦ AI payout suggests category meals Mar 6 16:20 finance payout category meals (accepted) Mar 6 16:21 finance payout payment_ref PX-2231
Each line is hash-chained to the one before. Your database keeps the current truth; Clipless keeps how you got there.
Seven real workflows. Open one.
Each runs in your browser, nothing to install. Click through it, send it back, skip to a deadline, and watch the record fill up beside the form. Or describe your own below.
An AI that reads the certificate
A supplier uploads a certificate. An AI stage reads it, fills in the standard and the expiry, flags the number when something is off and sends it back or on. At sign-off it suggests an answer; quality take it or don't.
Repeating sections
A tax return with as many dependents and businesses as the household has. Every item is validated on its own; the accountant sees them as tables with totals.
Send it back for changes
An expense claim the manager can return. It comes back as round 2, with notes on the amount that only finance can read.
Checks that take a while
A loan with four people, an identity check that finishes later, an AI risk read, generated PDFs and an email.
The same check, every year
An annual supplier review that starts from last year's answers and shows the reviewer exactly what changed.
Regulated intake
A patient intake with a one-time code, contact details kept out of caches, an encrypted history and consent that must be given.
Conditional stages
B2B onboarding where a risky industry or score routes the application through compliance.
It's just React.
The workflow is a file next to your screens, built with the components you already have. Deploy it with one command. Your backend starts sessions over HTTP and listens for signed webhooks; the engine does the rest.
npx clipless login npx clipless deploy ./expense-claim.tsx
Your frontend
- React
- Next.js
- React Router
- TanStack Start
- Inertia.js
- Astro
- Vite
- TypeScript
Headless hooks and fields that drop into the app you already have.
Your components
- Tailwind
- shadcn/ui
- Radix
- MUI
- Bootstrap
- Ant Design
- Your design system
Every field renders whatever you hand it, or start from ours.
Your models
- Claude
- OpenAI
- Gemini
- Groq
- OpenRouter
- Self-hosted
AI steps run on your own key, on any OpenAI-compatible API.
Your backend
- Node
- Bun
- Deno
- Express
- NestJS
- Django
- FastAPI
- Flask
- Laravel
- Rails
- Phoenix
- Spring
- .NET
- Go
Signed webhooks and an HTTP API. The engine is hosted; your servers just listen.
Pay per session. Every feature on every plan.
A session is one run of a workflow: one claim, one onboarding, one annual check. Test environments are free and unlimited. You bring your own keys for AI, identity checks and file storage, so there is no markup on providers.
- Unlimited test sessions
- 100 live sessions a month
- Every feature
- Community support
- Priced per live session; test sessions stay free
- Every feature, on every plan
- Your own keys for AI, identity checks and storage, with no markup
- Early-access teams help set the prices
Enterprise
A dedicated deployment with its own database and encryption keys, in the region you choose, plus single sign-on and custom terms such as multi-year retention, an SLA or a HIPAA business associate agreement, on request. Talk to us.
Clipless is in early access.
The engine isn't public yet. If you have a form that passes through more than one person or an AI agent, tell us about it, and we'll get you set up.