Open source · MIT · Agentic UI SDK
Your buttons, links and form fields become tools an agent can call. Your users just ask, in text or voice, right inside your app: nothing to install on their side.
Built into your app · Nothing to install for your users · Text and voice
npm i @owllayer/reactToday
Take one simple wish: a navy hoodie, size M, under €50. Menus, submenus, filters, sorting, pages: here is what it costs on an ordinary shop.
“A navy hoodie, size M, under €50.”
How it works
Take a component you already have. Add one declaration next to its action. The agent can now call it, and your code still does the work.
Your interface
€49
ShareWhat the agent receives
Live storyboard
A shopper wants a navy T-shirt, size M, under €50. With OwlLayer AI built into the shop, they say it once, hands free. The agent searches, filters and opens the product with the tools the current screen offers, and the list changes as the pages change.
T-shirts
6 results
OwlLayer Tee
€49
Harbor Tee
€65
Trail Tee
€45
Pocket Tee
€55
Crew Sweatshirt
€39
Fleece Jacket
€59
Shop assistant
Online
Available to the agent
App-wide
This screen
A demo shop. search_products, add_to_cart, select_shipping, select_payment and confirm_checkout, with their risk levels, come from the open-source React demo.
Paste an address. The scanner opens the page as a person sees it and lists the actions an agent could take there: search, add to cart, book, send. With the code to declare each one.
See it act
The open source demos, replayed as animations with their own tools. Beside each one, the protocol track shows what the agent does, step by step.
Search, add to cart and check out: the payment waits for approval.
Order #1042
Thank you, your order is confirmed.
Assistant
Find a navy hoodie and add it to my cart.
Check out.
Your order is confirmed.
Animated reconstruction of the demo, until its video recording.
Edit the catalog in plain language: deleting a product needs a yes.
Products
ProductStockPrice
Oak desk12€320
Desk lamp48€45€39
Linen cushion90€29
Wall shelf31€65
Assistant
Lower the desk lamp to €39.
Remove the discontinued chair.
Done, the catalog is up to date.
Animated reconstruction of the demo, until its video recording.
Plan a trip out loud: the booking is confirmed by the traveller.
Your trip
Lisbon 4 days
Flight · Hotel
Budget€1,140 / €1,200
Assistant
Four days in Lisbon, under €1,200.
Trip ready: 4 days, €1,140.
Book it.
Animated reconstruction of the demo, until its video recording.
Integration & models
Pick the integration that matches your product and the model that fits your use case. The same actions work across text, speech and realtime audio.
Your stack
OwlLayer AI Runtime
AITP · Policy · Approval · Context
Your model
Trust
Every action is an explicit capability of your application. Nothing runs that your code did not declare.
Four risk levels
Approval required
No DOM scraping
Agents receive structured contracts and the context you publish, never the rendered page.
Docs →Keys stay on the server
Provider credentials never ship in the browser bundle.
Docs →The server has the last word
Your server can block any tool, even one that a page has declared.
Docs →An approval the page cannot touch
The approval window renders in a closed Shadow DOM, out of reach of page scripts.
Docs →No. The agent is part of your web app: your users open your site and ask, in text or by voice. No extension, no desktop app, no AI subscription of their own, and no browser feature to turn on: it works today in Chrome, Edge and Firefox. Your server holds the model keys.
MCP declares tools on the server side: what your services can do. OwlLayer AI declares them on the interface side: your components and HTML elements become the actions available on the current screen. They work together, and the same agent can use both. MCP bridges are on the roadmap.
Yes. With the browser package, a few data-owllayer attributes on a button, a link or a field turn it into a tool, in plain HTML or a server-rendered page. React, Vue, Svelte and Angular packages follow the same model.
It can only call tools your interface declared, with input that passes their schema. High and critical tools pause for a human decision, your own handler performs the work, and your server can block any tool.
No. Wrap your app with the provider, then add tools one component at a time. Pages, routing and business logic stay as they are. To remove it, delete the provider and the tool declarations: there is no proprietary data format and nothing to migrate.
You do. The runtime runs on your server with your provider keys, and the agent only receives the context you choose to publish.
OwlLayer AI is in 0.x: the public packages are released and tested, and APIs can still change between minor versions. Pin your versions and read the changelog before upgrading.
Open source
Read the code, run the demos, open an issue, send a pull request. The roadmap and the discussions are public.
Start with one button and one action. Grow from there.
owlLaunch() · pending
OwlLayer AI is finishing its first public release. See you on the date below.
Tool resultowlLaunch()
{
"ok": false,
"status": "scheduled",
"opens_at": "2026-10-05T09:00:00.000Z"
}