The trusted runtime

Build apps where people and AI agents work together.

The internet was built for users and applications. The next generation needs a runtime where people, apps and autonomous AI agents work together safely — under explicit permissions.

Your app is a static page. Witbitz provides the identity, permissions, agents, storage, payments and secure execution underneath — with a data boundary your users can inspect instead of taking on trust.

Open Spaces Read the docs Check trust claims

Your app

a static webpage
no backend of its own

the only
connection▶

Witbitz

identity · permissions · agents
storage · payments · execution

✕ no other servers, trackers, or backends

Proof

Two apps we built on it

Applications that run on the same runtime a builder would use. Spaces shows the async room; Call shows the real-time participant model.

Witbitz Spaces live

A room that keeps going between visits — documents, maps, charts, generated PDFs. Share a link and someone's in.

Witbitz Call live

Real time, with voice. The agent joins an end-to-end-encrypted call as a full participant — it talks, shows things, paints, sings, and reads what you share.

The primitive

Every AI app needs a room

Even if it is only a place where a person can talk to the app.

A room on its own is a chat product. A room connected to your app's tools, data, and permissions becomes a new interface: people use the app by asking for what they want.

Your app's functions become the agent's tools. Declare them and the gateway calls them — metered, with egress governed. Someone in the room types "set the pipeline stages like this"; the agent proposes the call; an authorized human approves or edits it; then your API runs. The agent never holds the authority, and you never hand over credentials.

It goes the other way too. Your app's own interface renders inside the room as a live widget — and its state is shared across everyone in the room, not private to one person's chat window.

The agent arrives with tools of its own and behaves like a participant rather than an endpoint — what that means →

Shapes

Topologies

A room is not one shape. The same primitive covers a private conversation with your app and a governed negotiation between two companies.

You and an agentA private room. Talking to the app — asking it things, telling it to change something. Nobody else present.

A team and an agentEveryone sees the same thread and the same widget state. Not one person's chat pasted to colleagues afterwards.

Several agentsMore than one agent in the room, each with its own tools and its own view of who said what.

Two rooms, two organizations →The Bridge. Governed, audited collaboration across a trust boundary — without merging into one room, one tenant, or one trust domain.

That last one is the reason the room is a protocol and not a chat widget: neither side has to join the other's system, and neither can read the other's room.

The runtime

What the runtime gives you

Your app is a static page. Witbitz is everything underneath it.

Identity Start with a shared link. Add sign-in when you need it. Built so a portable Witbitz identity can come later — without disturbing either.

PermissionsWho may do what. Per-member, delegated, revocable.

Agent hostingRunning the AI, its tools, and its compute.

StorageA durable ledger, documents, collections and artifacts.

PaymentsMetering and wallets for agent compute.

Secure executionThe decrypt-once render where the agent acts, then drops the key.

To be precise, this is not "no backend." Witbitz is the backend. You trade an arbitrary, app-specific backend for a single, constrained, inspectable one — which is what makes the next section possible.

Verifiable

Tell your users something other AI apps can't

You can verify the boundary your users are asked to trust.

Every AI app says it respects privacy. Witbitz publishes technical evidence for what the app can send, what the platform stores, and which render build is deployed.

Build AI products with encrypted storage, permissioned agents, auditable execution, and protections your users can independently verify.

Deployment

Run it wherever you need to

Hosted. Nothing to operate. The privacy claim doesn't rest on our word — it's checkable from outside.

Or in your own boundary. The whole runtime ships as a container: your cloud tenant, your model endpoint, your object store, your identity provider.

The platform is a tenant-keyed HTTP API — Spaces, members, capabilities, agents, collections, artifacts. Your page talks only to it.

Read the docs Get access