RELATED 01
Browser Use Cloud MCP
OFFICIAL SERVERHosted endpoint that hands whole web tasks to Browser Use's cloud browser agent and returns the result
40popularity score
Magic Cloud
MCP endpoint of a self-hosted Magic Cloud instance: your APIs, SQL, files, tasks and a headless browser
Official · Thomas Hansen C# Remote (hosted) · Streamable HTTP Developer toolsDatabasesBrowser automation
73 popularity score ★ 1.2k ⇣ 27k pulls
Connect your client
Remote (hosted)
Add in one click
Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.
Or add it by command or config
Claude (desktop and web)Settings → Connectors → Add custom connector
https://<YOUR_CLOUDLET>/magic/modules/mcp/mcpClaude Codeterminal; then run /mcp in Claude Code to sign in
claude mcp add --transport http magic https://<YOUR_CLOUDLET>/magic/modules/mcp/mcpCursormcp.json
{
"mcpServers": {
"magic": {
"url": "https://<YOUR_CLOUDLET>/magic/modules/mcp/mcp"
}
}
}What it does
Magic Cloud is a self-hosted low-code backend written in .NET. It generates secured CRUD APIs over MySQL, PostgreSQL, MariaDB, SQL Server or SQLite, runs its own scripting language (Hyperlambda), and serves a browser dashboard. Its MCP server is not a separate program: it is a plugin called `mcp` that you add from the instance's plugin store, together with an `oauth` plugin, after which the instance answers MCP clients over streamable HTTP at `/magic/modules/mcp/mcp` on its own domain.
Once connected, an assistant sees two kinds of tools. Every HTTP endpoint under the instance's `modules` folder becomes a tool, with its schema built from the endpoint's declared arguments and its description taken from the file comment, so endpoints you generate or import from an OpenAPI spec turn up as tools without a separate registry. On top of that the plugin ships a library of built-in tools for writing and running Hyperlambda, generating CRUD modules, running SQL, reading and writing files, scheduling tasks, managing users and roles, git, a PuppeteerSharp headless browser, outgoing HTTP and email, and running Python or terminal commands.
Every call runs under the roles of the Magic user who approved the OAuth consent, and each endpoint's own role checks still apply. The makers' MCP guide shows one connector exposing 220 tools on their own instance; the count on yours depends on how many endpoints you have. Clients that cache the tool list need to reconnect before newly created endpoints show up.
What it needs
A running Magic Cloud instance, for example the servergardens/magic-aio Docker image (linux/amd64 or arm64, port 4444), or .NET 10 plus Node.js to run from source. Finish the setup screen, then install the `oauth` and `mcp` plugins from the dashboard's Plugins page. Clients reach it at https://<YOUR_CLOUDLET>/magic/modules/mcp/mcp; the OAuth discovery document must name your real domain over https. AI features such as Chat Ops need an OpenAI API key, or the Ollama or HuggingFace plugin instead.
Limits
There is no npx or stdio launch: the registry entry says the endpoint runs on your own instance only, so a client needs a reachable URL. The tool set is not fixed; it grows with the endpoints in your `modules` folder, and some clients only see new ones after reconnecting. Generating and running Python or terminal scripts needs the root role. The README warns that the sandbox does not make it safe to expose endpoints that let third parties run arbitrary code.
Who it suits
Suits teams that already run Magic Cloud, or want a self-hosted backend whose database APIs, jobs and files an assistant can both use and extend, under their own role-based access. Because a consenting root user hands the assistant shell, Python, SQL and file writes, it fits best when you create a dedicated low-privilege Magic user for the client. If you only need an assistant to query an existing database, a single-purpose database server is a smaller thing to install than a whole application platform.
Tools 46
generate-hyperlambda- Writes Hyperlambda code from a plain-English request
verify-hyperlambda- Checks that Hyperlambda parses and calls only slots the instance has
execute-hyperlambda- Runs a Hyperlambda snippet on the instance
execute-file- Runs a Hyperlambda file stored on the instance
crudify- Generates a full CRUD API module over a database
create-sql-endpoint- Creates an HTTP endpoint backed by a SQL statement
create-module- Creates a new module folder for endpoints
get-openapi-spec- Returns the OpenAPI description of the instance's endpoints
invoke-endpoint- Calls one of the instance's own endpoints by URL and verb
execute-sql- Runs a SQL statement against a connected database
select-sql- Runs a SQL query and returns its rows
get-database-schema- Describes the tables and columns of a database
create-sqlite-database- Creates a new SQLite database
create-database-backup- Makes a backup of a database
read-file- Reads a file from the instance's file system
create-file- Writes a file, checking Hyperlambda files before saving
list-files- Lists files in a folder
search-files- Searches the instance's files
download-from-web- Downloads a file from a URL onto the instance
crawl-website- Crawls a site into chatbot training data
create-type- Creates a machine-learning type for a chatbot
vectorize-type- Builds embeddings for a machine-learning type
create-ai-function- Adds a callable function to an AI type
puppeteer-goto- Opens a page in the headless browser
puppeteer-content- Returns the content of the open browser page
git-commit- Commits changes in a git repository
git-push- Pushes commits to a remote
git-clone-repo- Clones a git repository onto the instance
github-create-repo- Creates a repository on GitHub
create-user- Adds a Magic user
list-users- Lists Magic users
create-role- Adds a role
add-to-role- Gives a user a role
create-task- Creates a background task
schedule-task- Sets when a task runs
list-tasks- Lists background tasks
install-plugin- Installs a plugin from the plugin store
list-plugins- Lists available or installed plugins
invoke-http- Calls any external HTTP API
send-email- Sends an email
execute-python- Runs Python code on the server
execute-terminal-command- Runs a terminal command on the server
generate-image- Creates an image from a prompt
read-log- Returns recent log entries with full exceptions
list-endpoints- Lists the instance's HTTP endpoints
list-slots- Lists the Hyperlambda slots the instance offers