magic

Magic lets you describe a full‑stack application or an AI‑driven tool in plain English and receive a running system that includes a database, a secured API and a frontend. The platform is open source, self‑hosted and runs on a lightweight “Magic cloudlet” that you can start with a single Docker‑compose command. All generated code is compiled to a whitelisted abstract syntax tree that only executes operations you have permitted.

The tool is aimed at developers who prefer to own their infrastructure rather than rent cloud services. You can keep an existing SQL database, SQL Server, MySQL, PostgreSQL or SQLite, and wrap it in role‑gated endpoints without migrating data. For teams that need custom line‑of‑business software, the system can produce CRM, ERP, logistics or back‑office dashboards from a single sentence. The built‑in Chat Ops agent guides you through the required decisions (data model, authentication, UI style) and then creates the full stack on the same cloudlet, so there is no separate deployment step.

Using Magic starts with a prompt such as “Create a CO2 calculator that allows people to measure their CO2 impact.” The platform fetches a sandbox description from https://hyperlambda.dev/.well-known/hyperlambda-agent, loads any needed files with `io.file.load`, and executes them with `execute-file`. The result is a SQLite database, public and role‑gated API endpoints, and a single‑page application that is immediately accessible at the local address (e.g., http://localhost:5555). The default credentials are root/root, and you are advised to change the password before exposing the service. A 14‑day free trial is offered, and the source code is available under the MIT license on GitHub, where the project has accumulated over 1,200 stars.

Key features

• Generate database schema, API and frontend from a natural‑language description
• Support for four SQL engines and role‑gated API endpoints without data migration
• Built‑in Chat Ops agent that interacts to refine requirements and builds the stack automatically
• Self‑hosted deployment via Docker Compose, running on a single cloudlet
• Open‑source MIT‑licensed codebase with community contributions visible on GitHub
• AST‑based execution model that restricts runtime actions to a whitelist