Uses Pydantic models for type-safe, structured responses and automatic JSON schema generation from type hints when interacting with Redfish endpoints.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Redfish MCP Servercheck the system health and show me any critical alerts"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Redfish MCP Server
A Model Context Protocol (MCP) server that enables AI agents and LLMs to control Redfish-enabled hardware and simulators through a secure API. Built with FastMCP for enhanced development experience and structured output.
Features
This MCP server provides tools for:
Power Operations: Control system power states (On, Off, Restart, etc.)
System Inventory: Get detailed hardware information and specifications
Event Logs: Retrieve and manage system event logs (System, Security, Manager)
Health Monitoring: Monitor system health, status, and sensor readings
User Account Management: Manage BMC user accounts and permissions
Sensor Monitoring: Read temperature, fan, power, and voltage sensors
Configuration Management: Check connection status and auto-configure from environment
Architecture
FastMCP Framework: Built with FastMCP for better development experience
Structured Output: Uses Pydantic models for type-safe, structured responses
Environment Configuration: Auto-configures from environment variables
Secure Authentication: Username/password authentication with configurable SSL
Development Tools: Compatible with
mcp devmode and MCP inspector
Security
All communications with Redfish endpoints are authenticated using username/password credentials. SSL verification can be configured (disabled by default for self-signed certificates common in BMCs).
Installation
This project uses uv for dependency management:
Usage
Environment Variable Configuration (Recommended)
The simplest way to configure the Redfish MCP server is using environment variables. The server will automatically connect on startup if these are provided:
MCP Client Configuration
Configure the server in your MCP client with environment variables:
Claude Desktop
Add to ~/.config/claude-desktop/claude_desktop_config.json:
LM Studio
Add to your LM Studio configuration (mcp.json):
If developing under Windows with WSL2 where the server is running in WSL, you may need to set the REDFISH_HOST to the WSL IP address or use localhost if accessing from Windows:
Manual Configuration
If you prefer not to use environment variables, you can use the tools to configure the connection:
Use the to check current configuration status
Use the to set up connection parameters manually:
Host URL (e.g.,
https://192.168.1.100)Username and password for BMC authentication
SSL verification settings
Use the available tools to interact with your Redfish-enabled hardware:
Available Tools
redfish_get_config_status- Check current configuration status and connectionredfish_configure- Configure connection to Redfish endpoint (if not using env vars)redfish_get_system_info- Get system information and inventoryredfish_power_control- Control system power stateredfish_get_event_logs- Retrieve system event logsredfish_get_health_status- Get system health and statusredfish_manage_users- Manage user accountsredfish_get_sensors- Get sensor readingsredfish_clear_logs- Clear system event logs
Example Usage
Development
FastMCP Development Features
This server is built with FastMCP, providing enhanced development experience:
Development Mode: Use
mcp devfor hot reloading during developmentMCP Inspector: Compatible with MCP inspector for debugging and testing
Structured Output: All responses use Pydantic models for type safety
Auto-generated Schemas: Tools automatically generate JSON schemas from type hints
Running the Server for Development
The server requires a Redfish hardware or simulator to connect to. Ensure the Redfish endpoint is accessible and the credentials are correct.
You can use Redfish-Mockup-Server for testing with simulated Redfish endpoints. Get the mockup data from the DMTF Redfish Mockup bundle.
Unpack them, install dependencies and run the mockup server: