One prompt, one deploy
Most platforms start you with a form. Company name, billing address, a verification email, a project wizard, a region dropdown you have to research. Only then do you get to the part you actually wanted: your code, running somewhere, on a URL you can share.
Brainpod starts somewhere else. You are already in your editor, and there is already an agent sitting next to your code that knows what the project is. So that is where the deploy begins.
The prompt
Inside your project, you hand your agent one instruction:
Install the Brainpod skill from github.com/brainpodnl/skills and use it to
deploy my project to Brainpod eu-west-1.
Sign me up using the Brainpod skills, work out what the project needs, and
hand me the live URL when it's up.
That is the whole interface. It works in Claude, Cursor and Codex, and it works the same way in all three, because the intelligence lives in the skill rather than in any one editor.
What the agent actually does
The prompt looks like a wish, but underneath it is a sequence you can follow step by step. Five phases, in order.
Install
The agent pulls the Brainpod skill and creates your account. No form, no email round-trip, no password to invent. The account exists because the agent asked for one on your behalf.
Inspect
Then it reads the project the way a colleague would on their first day. A
next.config.ts means Next.js. A prisma/schema.prisma pointed at Postgres
means you need a database, not just a container. A package.json engines
field pins the Node version.
This is the step that makes the rest possible. Nothing is declared twice: your repository is already the source of truth about what your project needs, so the agent reads it instead of asking you to restate it in a dashboard.
Provision
Now it builds the stack it just inferred: a route, an app container, and a Postgres 17 instance attached to it. Wired together, patched, with backups and metrics already on.
If you would rather start from something known-good, the same step can pull a blueprint instead — a stack we maintain, provisioned whole.
Deploy
Your image gets built and pushed, TLS is issued, and the release starts up next to the one already running. Traffic moves over only once the new container reports healthy. Nobody sees a maintenance page, and there is no window where half your visitors get the old build and half get the new one.
Live
The agent hands back a URL. It resolves, it has a certificate, and it is
served from eu-west-1 in Ede, the Netherlands.
Why the region is not a dropdown
Everything above runs on Dutch infrastructure, under Dutch and European law, and we do not replicate or transfer your data outside the EU. That is not a setting you have to find and switch on — it is the only way Brainpod runs.
Data residency tends to be treated as a compliance checkbox, something you justify to a procurement team after the fact. We think it is closer to a property of the system, like uptime or latency: either it holds everywhere by construction, or you cannot really claim it.
Agents need somewhere safe to work
The honest reason we built the flow this way is that the thing deploying your code is increasingly not a person. It is an agent, working faster than you can review, and it needs somewhere it can build, deploy and scale without a human unblocking it at every step — and without the blast radius that normally implies.
That is the interesting constraint. Give an agent real infrastructure and it will use all of it. So the platform has to be the thing that holds the line: isolated environments, deploys that can always roll back, and limits that are enforced by the system rather than by the agent's good judgement.
A prompt is a small interface. What it opens onto is the part we care about.
Try it
Hand the prompt to your agent and watch it work. Every new account starts with a 14-day trial, and you do not need payment details to begin.