Deploy by conversation.
consostack plugs into your AI coding assistant. Tell it what to put online — a git repo, a plan, maybe a database — and it sets up the hosting, builds your app, and hands back a live URL. No Dockerfiles, no dashboard to learn.
For thirty years, hosting meant learning the host — consoles, config files, pipelines. A second craft to master, just to put the first one online.
That was always backwards. You don't want to operate infrastructure. You want your idea on the internet.
So consostack makes the interface language. You already describe software to an AI to build it — putting it live belongs in the same sentence. You talk; the platform does the operating.
You talk. Your AI ships it. You never touch a server.
Whether it's your first deploy or your hundredth, it's the same single move: tell your assistant what you want live, and it handles the account, the build, and the URL on consostack — no Dockerfile, no pipeline to babysit.
- Ask in plain language — it creates your account and picks a plan
- Your framework is detected from the repo; nothing to configure
- Managed Postgres & Redis, attached with one instruction
- A real URL with HTTPS, ready to share the moment it's up
- Drive it from MCP, the REST API, or a web console
- If a build breaks, you get told in a sentence — not a stack trace
Your assistant does the hosting. You just ask.
Here's the whole idea: consostack is a service your AI assistant can operate on your behalf, over MCP — the open standard that lets AI tools take real actions, not just chat. You connect it once. After that, putting something online is a sentence in the same chat you already code in.
Connect it once
Add one endpoint to any MCP-capable assistant — Claude, Cursor, and others. That's the entire setup, and you never repeat it.
Ask for a deploy
Point it at a public git repo and name a plan: “Deploy this on Pro, with a Postgres.” It shows you the price, creates your account, and takes the key — you fill in nothing.
It goes live
consostack clones the repo, detects the framework, builds it, and serves it behind HTTPS on your own subdomain — usually under a minute.
To start you need an AI assistant that speaks MCP (like Claude or Cursor) and a public git repo — or pick a ready-made template and skip the repo entirely. Once it's live, the same conversation keeps managing it: add a database, read the build logs, roll back, or point a custom domain at it.
The dashboard was never the point. Shipping was.
So stop operating your hosting and start describing it. Add the endpoint to your assistant, and put something live in your very next message.
You don't pick a plan. Your assistant sizes it.
Tell it what you're putting online and it quotes the plan that fits — then you just say yes. No pricing form, no checkout to click through. This is the menu it chooses from.
- 3 projects
- 256 MB · 0.5 vCPU
- 1 database
- 20 projects
- 1 GB · 1 vCPU
- 5 databases · 10 domains
- 50 projects
- 2 GB · 2 vCPU
- 20 databases · 50 domains
- 200 projects
- 8 GB · 4 vCPU
- 100 databases
Prices are indicative during preview. You never select a tier by hand — your assistant orders the right one, and you approve it in the same chat.