I Built an AI DevOps Engineer That Remembers Your Infrastructure
I want to tell you about something I've been building — not as a sales pitch, but as a developer explaining a technical problem I solved.
The Problem
When you're a solo developer or a small startup team, you handle your own infrastructure. That means when nginx throws a 502 or a container enters CrashLoopBackOff or your server runs out of memory at 11 PM — you're alone.
The tools available are:
- Google (fast, generic answers)
- Stack Overflow (accurate, often outdated)
- ChatGPT (helpful, but stateless — no memory of your setup)
None of these know your infrastructure. Every session, you start from scratch: "I'm running Node 18 on Ubuntu 22 with PostgreSQL 15 and Nginx..."
What I Built
ARIA is an AI senior DevOps engineer that specializes in exactly this. Here's what makes it different:
4-Week Memory
ARIA remembers your infrastructure across sessions. Your OS, your stack, your previous errors, what fixes you applied. When you come back with a new error, it already knows your context.
Structured Fix Format
Every response follows:
- Root Cause — what actually broke
- Commands — the exact fix commands (no placeholders)
- Verify — how to confirm the fix worked
- Prevent — how to stop it happening again
GitHub Repo Scanner
Connect a repo and get an A-F grade on:
- Security vulnerabilities
- CI/CD configuration quality
- Dependency health
- Docker best practices
Live Server Monitoring
Connect your server and get AI-powered diagnosis when metrics spike — CPU, RAM, Disk.
Who It's For
- Solo developers managing their own servers
- Small startup teams (2-10 people) without a dedicated DevOps engineer
- Junior to mid-level developers learning DevOps on the job
- CTOs who need to stay on top of infrastructure without spending all day on it
Pricing
- Free: 20 messages/day
- Pro: $10/month (unlimited)
- Team: $39/month (5 seats, shared context)
Try It
step2dev.com — no credit card required to start.
I'm the solo developer behind this. If you try it and have feedback, my email is on the site. I read every message.
I built ARIA to solve exactly this.
Try it free at step2dev.com — no credit card needed.
Top comments (4)
This is actually a very real problem.
Most AI tools today are powerful but stateless, which makes them less useful for infrastructure debugging where context matters a lot. The idea of persistent memory + structured output (root cause → fix → verify → prevent) is what makes this interesting.
Curious how you're handling context storage and drift over time especially when infra changes frequently.
Great question — this was actually one of the hardest parts.
Right now ARIA stores a lightweight infra profile (OS, stack, services, past errors + fixes). When logs, repo scans, or monitoring signals show something different, the context updates automatically.
Still improving how it detects infra drift since environments change a lot.
If you're curious to test it with a real setup, feel free to try it: step2dev.com — would love your feedback.
The 4-week memory window is a smart constraint — long enough to catch recurring issues, short enough to avoid stale context poisoning the diagnosis. How do you handle contradictory state when a server config changes mid-window?
Exactly — the 4-week window was chosen to balance useful memory without letting old context pollute diagnosis.
When ARIA detects contradictory signals (like different runtime versions or config changes), it refreshes that part of the infra profile or triggers a re-scan of the environment.
Still iterating on this because real infra changes quickly.
If you're interested, you can try it here: step2dev.com — I'd be curious to see how it behaves with different setups.