Hey,
A good automation should not start with a blank prompt, a cron expression, an MCP key, an AI provider key, and the question "what exactly should I make this thing do?". That is a lot of homework before the agent has done anything useful.
So we are testing Agent Profiles in Quave ONE.
An Agent Profile is a ready-to-run automation for a problem you probably already have. You choose the environments, turn it on, and Quave ONE creates the schedule, the limited MCP access, the notification behavior, and the first run. It is not a chatbot with a nice name. It is a real automation, with its runs and results visible in your account.
Every profile also creates its own MCP key, named after that profile and limited to exactly the tools it needs. So when you open the Audit page, it is easy to see what the Morning Health Check key did, what the VM Savings Advisor key did, and what neither key was allowed to do.
The important part is also what it cannot do. These first profiles are read-only. They inspect your infrastructure and tell you what to do. They do not deploy, scale, delete, change a setting, or touch a secret.
We started with two because I would rather learn from two useful ones than publish a gallery of twenty random AI ideas.
1. Morning Health Check
This one runs every morning at 8:00 AM in the timezone of the browser you use to enable it, over the environments you selected.
It looks for things that should make you uncomfortable before they become an outage: stopped workloads, error spikes, strange CPU or memory behavior, missing or stale backups, and production environments without enabled alerts.
When everything is healthy, it stays quiet. Nobody needs another "all good" notification every morning.
When something needs attention, it should lead with the action, for example: "Enable daily backups on prod-db (Environment → Backups)," then show the numbers that made it say that. The goal is a useful answer on your phone in 10 seconds, not a 2-page AI essay.
2. VM Savings Advisor
This one is for Quave ONE Connect accounts. It runs every Monday at 8:00 AM and looks at the previous 14 days of environment and VM usage.
Connect is the Quave ONE plan that runs the same platform in your own cloud account, so you keep and pay your cloud provider directly while we operate the platform with you. That is why this profile can look at your VMs too.
It first looks for safe reductions at the environment level: replicas, environment size, requests, and limits. It only recommends reducing a VM when the freed capacity makes that safe, and it keeps application capacity separate from data capacity.
It never changes anything. A low point in one graph is not a reason to remove capacity, so it needs sustained headroom over two weeks and it should show you the normal usage and the peak before suggesting anything.
Maybe you need both. Maybe one is useless for you. That is exactly what I want to learn.
We are going to maintain them as products, not prompts
These profiles will have versions.
We will improve the checks, the instructions, the schedule, and the allowed tools as we learn. When a new version is ready, your enabled profile will show that an upgrade is available and you will be able to apply it.
There is also a new account-level notification you can enable for Agent Profile updates. Turn it on in Account → Notifications if you want the account to tell you when a profile has a new version, instead of discovering it only when you open Automations.
You will not wake up one day with an automation that silently gained new permissions. If an upgrade changes the tools it can use, Quave ONE will ask you to review and enable it again. The profile is useful only if you can trust what it does and what it cannot do.
5 million tokens so you can actually test it
For this test, we are giving you 5 million tokens.
Use the profiles many, many times. Enable one, look at the first result, change the environment scope, run it again, try both, turn one off if it is noisy, upgrade it when we release a version. Do not preserve the tokens like they are a museum piece. The point is to learn whether these agents help you run your infrastructure better.
Then please reply to this email.
I want the good feedback and the bad feedback:
- Which profile did you try?
- Did it find something useful, or did it create noise?
- Was the notification clear enough to act on?
- What did you expect it to check that it did not check?
- What profile should we build next?
The last question is probably the most important one. If there is a boring thing you check every day, every Monday, before every deploy, or after every incident, tell us. That is probably a better Agent Profile than a feature we invented in a meeting.
Go to Automations in your Quave ONE dashboard, enable one of the two profiles, and reply with what happened. I read every response.
Have a great week and go enjoy the people you love.
Abraços,
Filipe