Ask for it, watch it ship.
A sandboxed container that writes the code, runs it, and puts the result live at a URL you can open.
Build me an inventory tracker with a CSV import, then put it online.
$ nemu computer start --tier standardcontainer up in 1.4s · 512 MB · node 22$ npm create vite@latest inventory -- --template react-tsadded 214 packages in 6s$ npm run builddist/index.html 0.51 kBdist/assets/index-4f2a.js 148.22 kB$ nemu publishlive at inventory.nemu.page
How big is the thing you want built?
Memory and cores are allocated from what the build actually asks for, so there is no size to pick. The only thing that changes with the work is how long the box stays up, and that is what the minute rate is for. The computer runs on credits, so it starts on Starter.
A Next.js app
A Next.js app
Install, migrate, build and ship. Big enough that the install is the slow part, not the model.
- Live at yours.nemu.page
- Bring your own database
- Time on the box
- about 25 minutes
- Credits it draws
- €0.50
- Performance
- High performance
- Plan
- Pro and up
Or point it at the accounts you already pay for.
Shipping to nemu.page is the fast path, not the only one. Connect your own database and your own host, and the computer builds against them, runs the migration and deploys where your product already lives.
Databases
- NeonServerless Postgres, branched per environment
- SupabasePostgres with auth and storage attached
- PlanetScaleMySQL, with the branch workflow
- UpstashRedis and a queue, over HTTP
Somewhere to deploy
- VercelPush the build straight into your project
- NetlifySites and functions, on your own team
- RailwayLong running services and workers
- RenderWeb services, cron jobs and disks
Credentials are stored as scoped connectors, the same way every other integration works, and the computer only sees the ones you grant it.
See every integrationThe part that usually stops at a snippet.
Asking a model for code leaves you holding a text file. The computer takes the same request all the way to something running at an address, and charges for the machine rather than the thinking.
A real container
Not a preview pane. A Linux sandbox with a filesystem, a package manager and a network, started in seconds.
It runs what it writes
Installs, builds and tests happen in the box. You see the output it saw, command by command.
Ships to a URL
One publish puts the build live at your own address on nemu.page, ready to send to somebody.
Billed by the minute
€0.02 a minute on Standard, €0.06 on High performance. The container stops when the work does.
Sized by the build
Memory and cores are allocated from what the install and the build actually need, so there is no plan to size up front.
Your models drive it
The model doing the thinking is one of yours, on your keys. Credits only pay for the machine.
Nothing to clean up
No local toolchain, no stray dependencies, no half configured runtime left on your laptop.
Start with your keys, keep the whole bill.
The gateway, the chat and the computer on one account. Free is enough to evaluate the gateway and the chat on your own keys, and your provider keeps billing the tokens exactly as it does today.