OME MCP — Dell OpenManage Enterprise MCP Server

A Model Context Protocol (MCP) server that exposes Dell OpenManage Enterprise (OME) management capabilities to AI agents via a Streaming HTTP transport. No direct iDRAC connections are made — all data flows through OME's REST API.

Features

Related MCP server: Redfish MCP Server

Architecture

AI Agent / MCP Client
        │
        │  HTTP POST /mcp  (Streaming HTTP transport)
        ▼
┌─────────────────────┐
│   ome_mcp_v5        │   Docker container (Ubuntu/Python 3.11)
│   FastMCP server    │
│   port 8000         │
└────────┬────────────┘
         │  HTTPS REST API calls
         ▼
┌─────────────────────┐
│  OME Appliance      │   https://192.168.1.145/api
│  (192.168.1.145)    │
└────────┬────────────┘
         │  OME manages iDRAC internally
         ▼
   Dell PowerEdge Servers

No connections are made directly to iDRAC — OME handles that internally.

Prerequisites

• Docker Engine 24+ and Docker Compose v2 on an Ubuntu host

• Network access from the host to https://192.168.1.145 (OME appliance)

• OME admin credentials

Quick Start

1. Clone / copy project files

2. Configure credentials

cd ome-mcp
cp .env.example .env
nano .env          # Set OME_USER, OME_PASSWORD

Never commit .env to version control.

3. Start container

docker compose up -d

The MCP endpoint will be available at http://<host>:8000/mcp.

TLS / SSL Toggle

No rebuild is required — just change the env var and restart.

Connecting an MCP Client

Point any MCP-compatible client at:

http://<host_ip>:8000/mcp

Example with mcp CLI:

mcp call --url http://192.168.1.200:8000/mcp ome_list_devices '{"top":10}'

Example with Python SDK:

from mcp.client.streamable_http import streamablehttp_client
from mcp import ClientSession

async with streamablehttp_client("http://192.168.1.200:8000/mcp") as (r, w, _):
    async with ClientSession(r, w) as session:
        await session.initialize()
        result = await session.call_tool("ome_list_devices", {"top": 5})
        print(result.content)

Pagination

All list tools accept top (page size, default 50) and skip (offset) parameters. Use the returned next_skip value to fetch subsequent pages:

{"top": 25, "skip": 0}   // first page
{"top": 25, "skip": 25}  // second page

OData Filtering

Pass standard OData $filter expressions via the filter parameter:

"Model eq 'PowerEdge R640'"
"Severity eq 'Critical'"
"LastRunStatus/Name eq 'Failed'"
"StatusType eq 'New'"

Security Notes

• Credentials are stored in .env (owner-read-only, not version-controlled).

• The server runs as a non-root user (mcpuser, UID 1000) inside the container.

• Credentials are never logged.

• For production, set OME_VERIFY_SSL=true and use a valid TLS certificate on OME.

• The MCP endpoint itself is plain HTTP on port 8000; place a reverse proxy (nginx/caddy) with TLS in front of it if external exposure is required.

Troubleshooting

Authentication errors:

• Verify OME_USER / OME_PASSWORD in .env

• Confirm OME is reachable: curl -k https://192.168.1.145/api/SessionService/Sessions

SSL errors:

• Set OME_VERIFY_SSL=false for self-signed certs

• Restart after changing: ./service_control.sh restart

Tools return empty results:

• Devices may not be discovered yet — use ome_create_discovery_job

• Check OME permissions for the API account