The core APIs for your apps.
Open your app to users, systems and AI agents.
Secure, with no lock-in.
The backend every app needs: sign-up and login, permissions, a MongoDB database, real-time data, payments, semantic search and MCP for AI agents, set up by you or your coding agents.
Free to start, no card.
No lock-in. Your data in a standard MongoDB, the open source RESTHeart engine underneath.
- UserSara, in the game
GET /matchesHer matches and trophies. Nobody else's.
- Systemthe match server, with an API key
POST /resultsRecords who won. It cannot touch an account.
- AI agentSara's coach, over MCP
"How did I do this week?""42 wins, 5 losses. Still number one."
The frontend is ready too
Whole apps with sign-up, login, teams and payments already working, for React, Angular and Vue, plus a shop with Stripe checkout. Clone one, point it at your service, and change what you want.
See the startersApps for people and for agents
An agent-native app has two kinds of users. People use its screens. AI agents use the same data over MCP. One backend, one set of permissions, for both.
People
Through the screens your app has
- Sign up and log in
- Read and write their data
- Pay, get emails, invite their team
AI agents
Through MCP, with a role you give them
- Claude, Cursor, or an agent of yours
- The same data, as tools
- Only what their role allows
People and AI agents use your app safely, with the permissions you set once: the same rules for both. How agents connect
Build it with your AI assistant
ChatGPT, Claude or Cursor write the app. Ulabase is the part they do not have to invent: accounts, payments, data. Give your assistant this prompt and it starts from an app that already works.
- 1 Sign up and create a free serviceIt takes a minute. No credit card.
- 2 Copy the prompt and say what your app doesOne line is enough to begin: "a booking app for a yoga studio".
- 3 Paste it into your assistantIt connects to our two MCP servers, the documentation and your service, starts from an app that already works, and tells you what to run.
I want to build an app on Ulabase (https://ulabase.com). Ulabase is a hosted, open source backend on MongoDB: user accounts and login, permissions, payments with Stripe, a MongoDB database with a REST API, and AI search. I do not want to write a backend of my own. My app: [describe your app here] Ulabase has two MCP servers. If you can use MCP servers, add both: - The documentation: https://d0c5a1.ulabase.app/mcp (public, no token). Read it before you write any code. - My service: https://[my service address]/mcp, with the header "Authorization: Bearer [an API key I give you]". Through it you read and write my data, and create collections and indexes, within what that key allows. If you cannot use MCP servers, read the documentation starting from https://ulabase.com/llms.txt Then: 1. Start from the React starter app, https://github.com/ulabase/starter-react, which already has sign-up, login and teams. 2. Write the backend configuration my app needs, its collections and permissions, in the file ulabase.setup.ts of the starter. 3. Build the pages of my app on top of the starter, with @ulabase/kit. 4. Tell me, step by step, what to run on my computer. I am not an experienced programmer. Ask me for my service address and for the API key when you need them: the address is on the Home page of my service on ulabase.com, the key on its MCP page.
- The documentation, as an MCP serverThe assistant reads the real API, the permissions and the kit while it works.
- Your service's own MCP serverWith a role you give it, it creates collections, indexes and the rest of the configuration, and reads the data.
- The Ulabase CLI, for coding assistantsClaude Code or Cursor write the configuration of the service as a file in your repository, and apply it with the command ulabase setup.
Teams that run on it
The engine under Ulabase is RESTHeart: open source, and in production since 2014.
Migration was incredibly smooth. Ulabase handled everything seamlessly. The support team was responsive and helpful throughout the process.
Ulabase saved us months of development time. We needed a reliable API layer for our MongoDB database, and it delivered exactly that.