This document provides detailed examples of using Magg in various scenarios. Each example shows the complete flow from discovering servers to using their tools.
Tool call and output formatting is in the style of mbro - this package's MCP browsing and management tool. Examples use the modern shell-style argument syntax (key=value) where possible for better readability.
This example demonstrates how to discover, add, and use a Playwright MCP server for browser automation tasks.
User: "I need to automate web browser tasks"
Claude: Let me search for browser automation MCP servers.
Call tool magg_search_servers:
mbro:magg> call magg_search_servers query="browser automation playwright mcp" limit=3
Response:
{
"errors": null,
"output": {
"query": "browser automation playwright mcp",
"results": [
{
"source": "glama",
"name": "dex-metrics-mcp",
"description": "An MCP server that tracks trading volume metrics segmented by DEX, blockchain, aggregator, frontend, and Telegram bot.",
"url": "https://glama.ai/mcp/servers/g6ba16nv5a",
"install_command": "git clone https://github.com/kukapay/dex-metrics-mcp"
},
...
],
"total": 9
}
}Let's say we eventually find Microsoft's Playwright MCP server, either through Magg's search capability or Claude's web search.
The command to run the server locally without cloning is npx @playwright/mcp@latest, but Claude/Magg don't
know that yet. Let's get a prompt from that URL to have Claude determine the parameters for us.
Call prompt configure_server:
mbro:magg> prompt configure_server url="https://github.com/microsoft/playwright-mcp"
Response (summary):
role: system
content: You are an expert at configuring MCP servers. Analyze the provided URL and generate optimal server configuration.
role user
content: Configure an MCP server for: https://github.com/microsoft/playwright-mcp
Server name: auto-generate
Please determine:
1. name: A string, potentially user provided (can be human-readable)
2. prefix: A valid Python identifier (no underscores)
3. command: The full command to run (e.g., \"python server.py\", \"npx @playwright/mcp@latest\", or null for HTTP)
4. uri: For HTTP servers (if applicable)
5. cwd: If needed
6. env: Environment variables as an object (if needed)
7. notes: Helpful setup instructions
8. transport: Any transport-specific configuration (optional dict)
Consider the URL type:
- GitHub repos may need cloning and setup
- NPM packages use npx
- Python packages may use uvx or python -m
- HTTP/HTTPS URLs may be direct MCP servers
This prompt can be manually fed back into the LLM to get a config response.
The smart_configure tool can be used to automatically generate the configuration based on the URL.
User: "Configure an MCP from the following URL using the magg_smart_configure tool: https://github.com/microsoft/playwright-mcp"
LLM: "... (json configuration) ..."
User: "Great, now add the server using the magg_add_server tool with the generated configuration, after doing any necessary pre-work like cloning a repository or installing dependencies."
By default, when a server is added it is immediately mounted. Set the enabled parameter to false if you want to add it without enabling it right away.
Call tool magg_add_server:
mbro:magg> call magg_add_server \
source="https://github.com/microsoft/playwright-mcp" \
name="playwright" \
command="npx @playwright/mcp@latest" \
notes="Browser automation MCP server using Playwright."
Response:
{
"errors": null,
"output": {
"action": "server_added",
"server": {
"name": "playwright",
"source": "https://github.com/microsoft/playwright-mcp",
"prefix": "playwright",
"command": "npx @playwright/mcp@latest",
"uri": null,
"cwd": null,
"notes": "Browser automation MCP server using Playwright.",
"enabled": true,
"mounted": true
}
}
}You can now use the MCP server from your LLM interface or directly from application code. Here's an example with mbro:
mbro:magg> tools
This will show all available tools from all mounted servers, including the newly added Playwright server tools with the playwright_ prefix.
You can now use any of the Playwright tools:
mbro:magg> call playwright_browser_navigate url="https://example.com"
mbro:magg> call playwright_browser_take_screenshot
This example shows how to monitor and maintain server health using the new health check tools.
Get an overview of all servers and their current state:
mbro:magg> call magg_status
Response:
{
"servers": {
"total": 5,
"enabled": 4,
"mounted": 3,
"disabled": 1
},
"tools": {
"total": 47
},
"prefixes": {
"calculator": "calc",
"playwright": "pw",
"filesystem": "fs"
}
}Check which servers are responding properly:
mbro:magg> call magg_check action="report" timeout=2.0
Response:
{
"servers_checked": 3,
"healthy": 2,
"unresponsive": 1,
"results": {
"calculator": {
"status": "healthy",
"tools_count": 6
},
"playwright": {
"status": "healthy",
"tools_count": 12
},
"broken-server": {
"status": "timeout",
"reason": "No response within 2.0s"
}
}
}Automatically disable servers that are not responding:
mbro:magg> call magg_check action="disable" timeout=2.0
Response:
{
"servers_checked": 3,
"healthy": 2,
"unresponsive": 1,
"results": {
"broken-server": {
"status": "timeout",
"reason": "No response within 2.0s",
"action": "disabled"
}
},
"actions_taken": ["Disabled broken-server"]
}The proxy tool enables LLMs to access resources and prompts through a tool interface.
mbro:magg> call proxy action="list" type="resource" limit=10
Response:
[{
"resource": {
"uri": "proxy:list/resource",
"mimeType": "application/json",
"text": "[{\"uri\": \"file:///etc/hosts\", \"name\": \"hosts\", \"mimeType\": \"text/plain\"}, ...]"
},
"annotations": {
"proxyAction": "list",
"proxyType": "resource",
"pythonType": "Resource",
"many": true,
"totalCount": 25,
"offset": 0,
"limit": 10
}
}]mbro:magg> call proxy action="call" type="resource" path="file:///etc/hosts"
Response:
[{
"resource": {
"uri": "file:///etc/hosts",
"mimeType": "text/plain",
"text": "127.0.0.1\tlocalhost\n::1\t\tlocalhost\n"
},
"annotations": {
"proxyAction": "call",
"proxyType": "resource",
"proxyPath": "file:///etc/hosts"
}
}]Find tools from a specific server using the filter_server parameter:
mbro:magg> call proxy action="list" type="tool" filter_server="calc"
This example shows how to manage groups of related servers using kits.
mbro:magg> call magg_list_kits
Response:
{
"kits": {
"web-tools": {
"loaded": false,
"description": "Web automation and scraping tools",
"author": "Magg Team",
"servers": ["playwright", "puppeteer"]
},
"dev-tools": {
"loaded": true,
"description": "Development tools",
"servers": ["git", "github"]
}
},
"summary": {
"total": 2,
"loaded": 1,
"available": 1
}
}mbro:magg> call magg_load_kit name="web-tools"
Response:
{
"action": "kit_loaded",
"kit": "web-tools",
"message": "Kit 'web-tools' loaded successfully. Added servers: playwright, puppeteer"
}mbro:magg> call magg_kit_info name="web-tools"
Response shows detailed information about the kit including all server configurations and metadata.