# Magic Cloud [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/polterguy/magic  
**GitHub Stars:** 1204  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/magic-cloud

## Description
Self-hosted low-code platform (Docker); its MCP endpoint runs on your own instance, not via npx.

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What the Magic Cloud MCP server does

The Magic Cloud MCP server connects MCP-compatible agents to a self-hosted Magic Cloud instance. After the MCP plugin is installed, HTTP endpoints in the cloudlet's `modules` folder become callable tools. This lets an agent invoke APIs created by Magic Cloud, including endpoints generated from databases, endpoints that wrap third-party OpenAPI services, and business logic written in Hyperlambda.

Magic Cloud is a low-code platform for building full-stack applications on your own hardware. Its surrounding platform can provide a database, secured REST API, authentication, role-based access control, scheduled jobs, and frontend files. The MCP layer gives agents access to the endpoints you choose to expose rather than requiring a separate hosted service.

This fits the 💻 Developer Tools category when you need an MCP connection to internal applications, legacy databases, or APIs that should remain under your control. It can also support database-driven agents that query or update application data through the platform's access controls.

## How it works

Magic Cloud runs as a Docker-based installation. The cloudlet hosts the application runtime and its modules, while the MCP plugin publishes the available HTTP endpoints to an MCP-capable client. The dashboard displays the cloudlet's MCP URL, which an agent uses to connect.

Endpoints can be created from existing MySQL, PostgreSQL, SQL Server, or MariaDB schemas, or generated by importing an OpenAPI or Swagger specification. Hyperlambda provides the application logic and runs in a sandbox. Its role-based controls can whitelist individual functions, allowing endpoint behavior to be restricted at execution time.

The platform does not require a deployment or publish step before saved code can be tested. Agents therefore work against the running self-hosted instance and the tools it currently exposes.

## Setup and configuration

Start the platform with Docker Compose:

```bash
curl -fsSL https://hyperlambda.dev/docker-compose.yaml | docker compose -f - up
```

Open `localhost:5555`, point the dashboard at `localhost:4444`, and sign in with the documented default credentials `root` / `root`. Install the `mcp` plugin, then configure Claude, Cursor, Codex, or another MCP client with the MCP URL shown by the cloudlet.

The MCP server is tied to your running Magic Cloud instance. The provided setup does not describe an npm package or an `npx` launch command, and no environment variables are specified in the project material.

## Tools and capabilities

The Magic Cloud MCP server can expose:

- Database-backed CRUD endpoints generated from existing schemas.
- Custom business logic and APIs written in Hyperlambda.
- Endpoints wrapping third-party services described by OpenAPI or Swagger documents.
- APIs used by database-driven agents to query and update application data.
- Scheduled or automated backend operations when those operations are represented by exposed endpoints.

The MCP plugin turns these HTTP endpoints into agent tools. Magic Cloud also includes dashboard features such as SQL Studio, the Endpoint Generator, Hyper IDE, Playground, user and role management, and a chatbot wizard, but the material specifically describes the MCP connection as exposing endpoints from the `modules` folder.

## Limitations and notes

Clients need MCP support and access to the cloudlet URL. The README reports an approximately 80% token reduction from its measurements, but that figure is an observed claim rather than a guaranteed result for every workload.

The platform's generated Hyperlambda can avoid nonexistent function calls by producing and checking an abstract syntax tree, but it can still produce logically incorrect code. Security and access behavior also depend on how endpoints, roles, and function permissions are configured.

_Full upstream README: https://allmcps.com/mcp/magic-cloud/readme_