Docs · Welcome

BuildMidas documentation

Agent teams that research, build, check, ship and improve your products — with you approving every step that matters.

Build somethingDescribe what you want, approve the plan, watch it get built and checked. Change what you gotEdit text, swap images, move blocks — or open a game's world in Studio. Put it livePublish on us, deploy to your own server, or ship to an app store. Let agents workResearch what to build, then keep it improving after launch.

What BuildMidas is

Most tools do one part of the job. A builder makes you something and stops. An SEO tool tells you what is wrong but cannot fix it. A research tool finds ideas but cannot build them.

BuildMidas is a set of agent teams that work on your products across the whole cycle: finding an opportunity worth pursuing, building the thing, checking it actually works, putting it live, and then improving it once real people are using it. You stay in charge at every step — nothing is published, deployed or charged without you approving it.

What you can build

Websites and web apps — from a one-page site to a database-backed app with its own server.
Games — 2D and true 3D, playable in the browser, with an editor for the world itself.
Mobile apps — installable Android builds you can ship to Google Play.
Desktop apps — for Windows, macOS and Linux.
Browser extensions — for Chrome and Firefox, shipped to their stores.

How a build actually goes

You describe it in plain words. A few sentences is enough. The assistant asks a couple of short questions when something genuinely matters, then writes a plan.
You approve the plan. Nothing is built until you do. You can change your mind, ask for something different, or start over.
Agents build it. You watch it happen — the steps being taken, the assets being made, and a live view of the work as it renders.
It gets checked. Automated review runs the thing in a real browser, plays it if it is a game, and looks for broken links, security issues and accessibility problems. Whatever it finds gets fixed and re-checked until it comes back clean.
You look at it and iterate. Preview it, edit it directly, or ask for changes in chat.
You put it live. Publish it on us, deploy to your own server, or ship it to a store.

Builds take a while — commonly 30 minutes to a couple of hours, longer for games. That is mostly the checking loop, which keeps fixing and re-testing rather than handing you something that merely looks finished. You can close the tab; the work continues.

Where things live

You want to…	Go to
Start or continue a build	AI Builder, with your past sessions listed underneath it
See everything you have made	My Builds
See what is live and where	Shipped and Published sites
Edit a game's world	Editor in the top bar, or the Edit button on the game
Find something worth building	Workload → Discovery
Teach the agents	AI Center
Connect Search Console, Analytics, stores, payments	Settings

What it costs you in credits

Work that runs an agent on your behalf uses credits: builds, assistant conversations, generated images and video, and agent runs. Reading your own dashboards, editing a build you already have, and browsing what you shipped do not.

If you ask an agent to do something that spends money or changes something publicly — a build, a deploy, a store submission, applying a change to a live site — you get a confirmation naming exactly what will happen before anything runs.

A good first hour

Build one small, real thing — a landing page for something you actually care about.
Preview it, then edit a headline and swap an image directly on the page.
Publish it and open the live URL.
Connect Search Console and Analytics so the agents can see how it does.

Building something →Agent teams & Discovery →