The OpenStack API and AI agents on Noiz Cloud, explained
The OpenStack API is coming to Noiz Cloud. It is the standard programming interface of OpenStack, the open-source cloud platform. In general, it lets people, scripts and AI agents create and manage cloud servers, snapshots and networks, with credentials limited to exactly what each one needs. This page explains how that works and how the safety model works, so that you can plan for it. It is not available yet, there is no date, and this page will change when it is.
Last reviewed: 19 October 2026, against the current OpenStack documentation. Nothing on this page is sold yet.
Official Documentation Reference
- OpenStack Keystone: Application credentials: credentials for programs instead of people.
- OpenStackClient: the command-line tool for the API.
- OpenStack Compute API: Create server: the call that starts a server.
What you can do today
Noiz Cloud servers are available now as self-managed Linux servers with root over SSH, listed on the Cloud page. An AI coding tool can already work on a Cloud server over SSH with a key of its own, and you can run your own models on one. What the API adds is control of the cloud itself, not only of what runs on a server.
What an OpenStack API lets an agent do
In OpenStack generally, these are calls a program can make:
- start a server of a chosen size from an image, with your SSH key;
- take a snapshot before a change and restore it afterwards;
- add a disk, a network or a firewall rule;
- list what is running and what it uses;
- stop and remove servers that are no longer needed.
An AI agent connected to an OpenStack API could, for example, take a snapshot of a server before it upgrades the software on it, or start a test server for a branch of your code and remove it when the tests pass.
How an agent's access can be limited
OpenStack has a feature made for this: application credentials. An application credential is an ID and a secret that a program uses instead of a person's password. Its documented properties are what make it safe to give one to an agent:
| Property | What it means for an agent |
|---|---|
| Scoped to one project | The agent sees only the servers and resources in that project |
| A subset of your roles | You can give it fewer permissions than you have yourself |
| Access rules | You can allow only named actions, for example listing servers and taking snapshots |
| Expiry | It stops working on a date you set |
| Cannot make more credentials | By default it cannot create or delete other application credentials |
| Revocable | Deleting it ends the agent's access at once |
The rules from the rest of this series still apply: one credential per agent, the least access that does the job, a person approving anything that deletes or costs money, and the credential withdrawn when the work is done.
How an agent connects
OpenStack tools read their settings from a clouds.yaml file on the computer that runs them, which names the cloud's address and the application credential. Command-line tools, scripts and AI agents that know OpenStack use that file. When the API is available for Noiz Cloud, we will publish the details here.
What to do now
- Plan your projects. Think about which servers belong together and which agent or person needs to reach each group.
- Use SSH keys of their own for every tool that works on your Cloud servers today, so that the habit is in place.
- Tell us what you would automate.
Common questions
- When will it be available? We have no date to give. This page will change when it is ready.
- Does it work with shared hosting? No. The OpenStack API is for Noiz Cloud. Shared hosting is managed in Plesk.
Talk to us
If you are planning infrastructure that agents will run, or want to be told when the API is ready, get in touch and tell us what you would like to automate.
