Meraki Magic MCP

# Meraki Magic MCP - Dynamic Full API This **hybrid dynamic MCP** provides access to **ALL** Meraki SDK methods (~804+ endpoints) through a combination of pre-registered common tools and a generic API caller. ## What This Gives You Full access to all Meraki API methods across all SDK sections: - **organizations**: ~172 methods - **networks**: ~114 methods - **switch**: ~101 methods - **appliance**: ~130 methods - **wireless**: ~105 methods - **camera**: ~45 methods - **devices**: ~27 methods - **sm**: ~49 methods - **cellularGateway**: ~24 methods - **sensor**: ~18 methods - **insight**: ~7 methods - **licensing**: ~8 methods - **administered**: ~4 methods **Total: ~804 methods accessible!** ## Architecture: Hybrid Approach To avoid MCP conversation length limits, this uses a hybrid approach: ### 1. Pre-Registered Tools (12 most common) Fast, convenient access to frequently-used operations: - `getOrganizations` - `getOrganizationAdmins` - `getOrganizationNetworks` - `getOrganizationDevices` - `getNetwork` - `getNetworkClients` - `getNetworkEvents` - `getNetworkDevices` - `getDevice` - `getNetworkWirelessSsids` - `getDeviceSwitchPorts` - `updateDeviceSwitchPort` ### 2. Generic API Caller (all 804+ methods) Access any Meraki API method via `call_meraki_api`: ```python call_meraki_api( section="appliance", method="getNetworkApplianceFirewallL3FirewallRules", parameters={"networkId": "L_123456"} ) ``` **Claude automatically routes to the right tool:** - Common operations → Pre-registered tool (faster) - Everything else → `call_meraki_api` (full coverage) ## Files - `meraki-mcp-dynamic.py` - Hybrid MCP server (12 pre-registered + generic caller) - `meraki-mcp.py` - Original manual MCP server (40 hand-coded endpoints) ## Helper Tools Included Discovery tools to find and use any API method: 1. **list_all_methods** - Browse all 804+ methods (optionally by section) 2. **search_methods** - Search for methods by keyword 3. **get_method_info** - Get parameter details for any method 4. **get_mcp_config** - View MCP configuration 5. **cache_stats** - View caching statistics 6. **cache_clear** - Clear response cache ## Usage with Claude Desktop Update your `claude_desktop_config.json`: ```json { "mcpServers": { "Meraki_Magic_MCP_Dynamic": { "command": "/Users/apavlock/meraki-magic-mcp/.venv/bin/fastmcp", "args": [ "run", "/Users/apavlock/meraki-magic-mcp/meraki-mcp-dynamic.py" ] } } } ``` ## Example Usage in Claude ### Discover available tools ``` Use the list_available_tools to show me all wireless tools ``` ### Search for specific functionality ``` Use search_tools to find all tools related to "firewall" ``` ### Get parameter details ``` Use get_tool_info to show me the parameters for organizations_getOrganization ``` ### Call actual API methods ``` Use organizations_getOrganizations to list all organizations ``` ``` Use networks_getNetworkClients with networkId="L_12345" and timespan=86400 ``` ## Advantages of Dynamic Approach ✅ **Complete Coverage** - All 800+ endpoints available immediately ✅ **Auto-Updated** - When Meraki SDK updates, you get new endpoints automatically ✅ **No Manual Coding** - No need to hand-code each endpoint ✅ **Consistent Interface** - All tools follow the same pattern ✅ **Error Handling** - Built-in error handling for all methods ✅ **Async Support** - All methods wrapped in async handlers ## Disadvantages vs. Manual Approach ⚠️ **Less Type Safety** - No Pydantic schemas for validation ⚠️ **Generic Descriptions** - Tool descriptions are auto-generated ⚠️ **No Custom Logic** - Can't add custom business logic per endpoint ⚠️ **Larger Tool List** - 800+ tools may be overwhelming in Claude UI ## When to Use Which Version **Use Dynamic (`meraki-mcp-dynamic.py`) when:** - You need access to many different API endpoints - You want complete API coverage - You're exploring/prototyping - You want automatic updates when SDK changes **Use Manual (`meraki-mcp.py`) when:** - You only need specific, well-defined endpoints - You want strong type validation (Pydantic schemas) - You need custom business logic - You want clean, curated tool descriptions ## Hybrid Approach You can use BOTH! Configure both servers in your Claude Desktop config: ```json { "mcpServers": { "Meraki_Curated": { "command": "/Users/apavlock/meraki-magic-mcp/.venv/bin/fastmcp", "args": ["run", "/Users/apavlock/meraki-magic-mcp/meraki-mcp.py"] }, "Meraki_Full_API": { "command": "/Users/apavlock/meraki-magic-mcp/.venv/bin/fastmcp", "args": ["run", "/Users/apavlock/meraki-magic-mcp/meraki-mcp-dynamic.py"] } } } ``` This gives you: - Curated, well-documented tools for common operations - Full API access for everything else ## Testing Before testing with your production API key, I recommend: 1. **Test tool listing** (doesn't call Meraki API): - Use `list_available_tools` helper - Use `search_tools` helper - Use `get_tool_info` helper 2. **Test with read-only operations first**: - `organizations_getOrganizations` - `organizations_getOrganization` - `networks_getNetwork` 3. **Avoid destructive operations until confident**: - Anything with `delete` in the name - Anything with `update` or `create` ## Error Handling All tools include built-in error handling that returns JSON with error details: ```json { "error": "Meraki API Error", "message": "Invalid API key", "status": 401 } ``` ## Auto-fills Organization ID If a method requires `organizationId` and you haven't provided it, the tool will automatically use your `MERAKI_ORG_ID` from `.env` file. ## Next Steps 1. Restart Claude Desktop (if config changed) 2. Test the helper tools first (they don't call Meraki API) 3. Test read-only operations with your production environment 4. Explore the full API at your own pace!