Apps

Your AI can already make you a working tool in an afternoon. The problem is where it goes. It runs once on the laptop it was made on, or it ships to somebody's personal Vercel account, and two weeks later nobody can find the link. That is what Apps fix. An app your agent makes on Runwork is a real full-stack application (React frontend, TypeScript backend, its own database) with an address, a team behind it, and someone able to open it on Monday.

It starts in whatever AI tool you already use. You tell your agent what you need: "our clients should be able to see their invoices, pull a report, and file a ticket." Your agent reads your workspace first, your existing records and your 3,300+ integrations, so what it makes fits the way your company actually works.

Then it works, and you watch. Your agent plans the pages and the data flow, writes the code, and runs it in a sandbox with live preview and hot reload, so you see the thing take shape without installing anything. For heavy work like crunching files or long research, it can hand off to an agent sandbox. Underneath, runwork init, runwork dev and runwork deploy are running for it. You never type them.

The app is wired into your team's data on day one. If your workspace already knows your customers, the app shows and edits those same records with no extra setup. If Stripe is connected, payment history is there. Nothing is siloed, so your customer portal, your sales dashboard and your support tool are all looking at the same numbers.

Shipping is the part that makes it the team's. One deploy and it is live worldwide with a URL you can paste in a message. Roles and groups decide who opens it and who can change it. Every change is a commit, so you can read the code and roll back through deployments at any point. And when finance asks for one more column six weeks later, your agent picks the same app back up, ships the change, and everyone has it. Nobody starts over.