Describe a business. Get a running system.
Dataceen turns a description of a business into a running system: a data model with its databases, an API, its own MCP server, SDKs, example data and a hosted application, built from your AI assistant.
From words to a system.
Say what the business does, in a sentence or a page, with documents, images or a database schema if you have them. Claude drafts the model; you see it as a picture and change it in conversation.
Everything generated, and it follows the model.
A database per model, a GraphQL API, search, events, SDKs in six languages, documentation and an MCP server of its own. Change the model later and all of it follows.
An application, and the code is yours.
Dataceen can build and host a dashboard, a site or a task program on it. The code lives in a private GitHub repository you can be added to or take over.
Connect
Server address
https://mcp.dataceen.com/mcp/studioClaude (claude.ai, Desktop, mobile)
Settings > Connectors > Add custom connector, then paste the address.
Claude Code
Add the server with this command, then run /mcp to sign in.
claude mcp add --transport http dataceen https://mcp.dataceen.com/mcp/studioCursor, VS Code and other hosts
Hosts that speak streamable HTTP use the same address. Hosts that only speak stdio:
npx -y mcp-remote https://mcp.dataceen.com/mcp/studioThe first screen offers two ways in
Sign in.
A free account takes a minute.
Continue as guest.
No account. You can draft models; drafts are kept 30 days. When the model is right, a link signs you up and makes it yours, and the same connection goes on working for you.
A first prompt
Create a model for a field-service company: customers, sites, elevators under service agreements, technicians with work orders, safety certificates that expire, monthly invoices.
What happens next
- 1The assistant shows what it will build from and drafts the model when you say so (a few minutes).
- 2You see the model as a picture and ask for changes, as many rounds as you like.
- 3When you say it is right, Dataceen creates the system.
- 4The assistant asks about example data, then which consumer a scope is for, and publishes it: that scope gets its own MCP server and API.
- 5If you want an application, the assistant asks three short questions and Dataceen builds and hosts it.
The tools
| Tool | What it does |
|---|---|
| about_dataceen | Facts about Dataceen, for the assistant to explain in its own wordsReads only |
| create_model | Starts a model with a name; your description is its first source |
| add_source | Adds text, an OpenAPI spec, a database schema, an AsyncAPI document, a document or an image |
| remove_source | Takes one source off the list |
| build_model | Drafts the model from the sources; creates nothing |
| get_model | The model: entities, relationships, scopes and their addresses, a picture linkReads only |
| update_model | Changes a created model: entities, properties, relationships |
| create_system | Creates the reviewed model as a running systemAsks first |
| generate_example_data | Fills the system with realistic records |
| update_scope | Decides what a consumer sees and may change |
| publish | Makes a scope live as its own MCP server and APIAsks first |
| get_sdk | The scope's SDK in the language you program in, with its guidesReads only |
| build_application | Dataceen builds and hosts a dashboard, a site or a task program |
| get_application | The application's state, address and codeReads only |
| change_application | Changes the live application from one request |
| take_down_application | Takes an application offline; its code staysAsks first |
| share_repository | Invites a GitHub user to the model's private repository |
| transfer_repository | Hands the repository over to a GitHub user or organisationAsks first |
| get_balance | Your credits and what builds are estimated atReads only |
A guest sees fewer tools: the ones for drafting a model, plus the link that makes it yours.
Costs and limits
- AI work is paid in credits; one credit is one euro cent. Every build shows its estimate before it starts, and a failed run costs nothing. See pricing.
- New accounts get credits to start, and the account's first system, with its first application, needs no card.
- Long steps report progress: drafting a model takes a few minutes, example data a few more, an application ten to forty minutes.
Data, security and support
- Sign-in runs through Dataceen's own authorization server (OAuth with Microsoft Entra External ID). Every call is counted and logged with who made it.
- Data is stored in the EU. See the privacy policy and the subprocessors.
- The code is yours, in a private GitHub repository you can take over.
- To disconnect, remove the connector in your assistant. To delete a model, use Studio or ask the assistant.
- Support: support@dataceen.com