OpenStack MCP Server is an MCP-based resource request service that provides API interfaces for requesting compute resources, storage, networking, images, and other resources from OpenStack cloud platforms. It runs as a high-performance asynchronous HTTP service built on Starlette and Uvicorn, with SSE streaming output for real-time feedback as requests are processed.

Use it when you want an AI assistant to query the current status of resources in an OpenStack cluster in real time, for example checking VM instances with a specific name filter and detailed verbosity, instead of using the OpenStack CLI or dashboard directly. It requires a running OpenStack environment along with admin credentials, an auth URL, username, and password, supplied as command-line parameters when starting the server, so it's a fit for teams already operating OpenStack infrastructure rather than a general cloud-resource browser.

get_instances retrieves OpenStack VM instances, with filtering and a verbosity level control supporting basic, detailed, and comprehensive information display. Beyond instance retrieval, the server is built for multi-dimensional resource requests spanning compute capacity, storage, networking, and images, with flexible filtering by name, ID, and other conditions, over a standard MCP interface for seamless integration with large language models.

Install via pip:

pip install openstack-mcp-server

Or from source:

git clone https://github.com/wangshqly0407/openstack-mcp-server.git
cd openstack-mcp-server
uv venv
source .venv/bin/activate
uv sync
uv run ./src/mcp_openstack_http/server.py --port 8000 --log-level INFO --auth-url 'http://<OpenStack-API-Endpoint>:5000/v3' --username '<OpenStack-Admin-User>' --password '<OpenStack-Admin-Password>'

Requires Python 3.10+ and an existing OpenStack environment. The service runs on port 8000 by default, with the MCP interface available at http://localhost:8000/openstack, and supports command-line parameters for the port, logging level, auth URL, username, and password.

Teams running OpenStack cloud infrastructure who want an AI assistant to query compute, storage, networking, and image resources directly through a standard MCP interface.