Skip to content
All posts

From idea to agentic app, in minutes

Why we built Something, what happens in the two minutes after you press enter, and where agentic apps go from here.

Shinjan PatraCo-founder & CTOAugust 6, 20261 min read

Most software never gets built. Not because it's hard — because between the idea and the running thing sits a queue: a developer to brief, a sprint to wait for, glue work nobody owns.

We built Something to delete that queue. You describe the agent you want in a sentence, and a few minutes later it's deployed — UI, backend, integrations, hosting, the lot.

What actually happens when you press enter

The builder reads your prompt and drafts a spec: what the agent does, which skills it needs, what its interface looks like. Then it writes the app, wires the connections, and ships it to its own subdomain.

  • Skills, not plumbing. Email, Slack, payments, voice calls — each is a capability the builder composes, not an API you integrate.
  • Your data, isolated. Every agent gets its own database branch. No shared tables, no noisy neighbours.
  • Editable forever. Come back, say what should change, and the builder edits the running app in place.

A concrete example

Here's the entire "brief me every morning" agent, as the builder sees it:

const agent = {
  name: "founder-brief",
  schedule: "every weekday at 7am",
  skills: ["gmail", "slack", "stripe"],
  output: "one short brief, priorities first",
};

That's not pseudocode for effect — specs really are this small, because the platform owns the hard parts.

The fastest teams we know don't write more code. They describe more precisely.

Where this goes

TodayNext
Text + voice agentsMulti-agent teams
20+ skillsMarketplace
Web + phoneNative mobile

We're letting people in batch by batch. If you want the queue deleted for your team, the beta form takes a minute.

Build the thing you just read about

Start with a sentence. Leave with a working agentic app.

From idea to agentic app, in minutes · Something