Trontra Agent & Apps building AI platform

What Trontra is: You describe what you want in plain language, and Trontra builds either a full-stack web app or a conversational AI agent. Both builders plan the work, build it in phases on a live timeline, check every change automatically, and keep every version so you can restore it. Trontra is in private beta. The agent builder is open to every workspace, and the app builder is in alpha for invited workspaces only.

How a build works: you write a prompt and choose Auto or Plan mode. You then get a structured plan, and the app builds in a chat with a live preview next to it.

Checks: every version must pass install, type-check, lint, unit tests and a production build. When a check fails, the builder fixes it itself, for up to three rounds.

Before publishing: a Launch Readiness check scores the app out of 100 across 11 areas. You can publish once every blocking check passes, and the code goes to your GitHub.

Live today: planning, building by chat, live preview, pinned comments, versions, ready-made modules, Stripe payments, Resend email, GitHub export, connections and secrets, and spend caps.

Paid plans: publishing and custom domains, team access, developer mode, and agents inside apps.

In development: moving an app to your own hosting.

Not built yet: a managed database, automatic database changes, buying domains, one-click Stripe connection, and a full in-browser editor with a terminal.

How a build works: you write a brief or pick one of six starter ideas, and choose a tone. Plan mode asks up to six questions first. The agent then builds in phases of up to three capabilities each, with test conversations for each phase.

Going live: a version goes live by itself only when at least 80% of tests pass and no boundary test fails.

Live today: create, plan, build, test, the playground, editing instructions and settings, versions, team roles, and spend tracking.

Paid plans: knowledge (reference text the agent answers from) and channels (hosted chat page, website widget, iframe embed, REST API).

Early access: agents inside apps.

In development: the visual canvas.

Not built yet: editing on the canvas, outside tools like webhooks, knowledge from URLs or files, and a history of past conversations. Agents have no schedules or triggers.