Operations
Once your software is running in the cloud, ACE helps you run and maintain it. Operations is the DevOps half of the platform: watch the live system, react when something breaks, and keep it cheap and secure. It applies the Software Factory AI model to infrastructure instead of features.
This pillar only fully applies to cloud-hosted apps. It needs a connected cloud account and discovered infrastructure.
What Operations covers
- Map the infrastructure of your code. From the connected cloud account, ACE builds an inventory of what's actually running and maps it to your codebase (the same discovery step described in Feature Development and Learn Path).
- Monitor the health of your services. Every confirmed resource gets a real health check that runs against it, so you see true up/down, not just what the provider claims.
- Monitor incidents. Problems surface automatically, grouped and de-duplicated so one real issue is one incident, not a flood of alerts.
- Read logs and check alarms. Investigate what your services are doing. Alarms from your monitoring tools flow in as incidents.
- Review the security of your cloud infra. Recurring, automated security reviews of your account.
The three operational surfaces
Monitoring & Incidents
- Health checks are real commands run against each resource. The verdict is decided by the command's exit code, not guessed by the AI. If a check fails, the AI writes an explanation of why.
- Incidents are created automatically from three sources: a failed health check, an alarm from a connected tool, or a signal from your deployment pipeline. You never create them by hand. ACE groups repeats under one incident (the "×N occurrences" badge), flags flapping, and settles an incident to resolved only after a grace period so a blip doesn't bounce.
- Shortly after an incident is detected, ACE analyzes it and writes a diagnosis into the incident's thread.
Costs
Recurring, automated reviews of your cloud spend. You set a schedule and a prompt (or a preset). The AI reviews your bill on a repeating basis and hands back a plain-language report flagging waste, anomalies, and unexpected growth, and ways to reduce cost.
Security
The same as Costs, but for security posture. Scheduled reviews of your cloud account produce a written report of findings on a repeating basis.
Cost and Security checks produce a written finding, not an automatic fix. To act on a finding, you discuss it with ACE and turn it into work (see below).
ACE interaction within Operations
Operations is not just dashboards you read. It is a place where ACE actively does the analysis work. Two kinds:
| Interaction | What ACE does | Where |
|---|---|---|
| Log review | Reads and reasons over logs and signals to diagnose a problem | Incident "Agent analysis"; Discuss in chat on an incident |
| Cluster verification | Runs real commands against live resources to confirm they're healthy | Health checks on discovered resources |
Reviewing a pull request is a Feature Development interaction, not an Operations one. The feedback loop below is where the two connect.
The connective tissue is Discuss in chat. On any incident, cost report, or security report, you can open a conversation already loaded with that item's context and analysis, and ask the AI follow-up questions: "why is this failing," "how do I reduce this cost," "is this finding real."
The feedback loop: Operations to Feature Development
Operations closes back into the Software Factory. When an incident needs a code fix, you click Create issue on it. That:
- creates a card on the Development board,
- carries the incident's analysis thread over as the card's conversation, and
- dispatches the AI immediately. The card starts In Progress, and the inner loop runs to produce a fix and, ultimately, a pull request.
So a failing service can become a diagnosed incident, then a dispatched fix, then a reviewed pull request, all in one platform. This human-triggered "Create issue" is the bridge. There is no fully automatic "detect and fix." A person always decides to turn an incident into work.
What it needs
- A connected cloud account (read-only) and discovered infrastructure. Without them there are no resources to monitor and no cost/security data to review.
- Alarms from external tools can flow in as incidents even without cloud discovery, via a webhook.
- A connected communication channel (Slack or Teams) for Cost and Security reports to actually reach someone. Without one, a report still runs and lands in the in-app inbox, but nothing gets pushed to chat. See Communications.