
sudiviz — X-Ray Vision for Cloud Infra | AWS WAUG Recap
Recap of my AWS London WAUG talk on sudiviz — an open-source Python CLI that maps, diagnoses, and fixes messy AWS accounts. Think Paris metro, but for VPCs.
Last Thursday I presented sudiviz — the open-source project I've been quietly raising in my spare time — at the Amazon Holborn office as part of AWS London Well-Architected User Group July Meetup .
Demoing a tool that critiques AWS infrastructure, inside an AWS office,needs special kind of braveness :D 
I had the perfect context to go right after Alan's excellent session on Infrastructure as Code reviews.

We all know what happened when us-east-1 went down & what happens when Pager kicks in at 2AM.

According to Gartner, 80% of the cloud outages come from misconfigurations — not code, but configurations and orphaned resources.
AWS networking is basically the Paris metro
Here's the mental model I used. If you've ever tried to read the Paris metro map for the first time, you know the feeling. Fourteen numbered lines, a few RER letters, connections that look sensible until you actually try to change trains, and stations that are technically the same stop but require a five-minute underground walk. It works. It's beautifully engineered. It's also a lot.
AWS networking is that map. VPCs, subnets, route tables, NAT gateways, ALBs, target groups, security groups with 47 rules each, ingress from this thing, egress to that thing, one lambda quietly reaching into an RDS in a different AZ. Every piece is documented. Every piece makes sense on its own. Put them together and you've got spaghetti that even the person who built it can't fully hold in their head six months later.
That's the problem sudiviz was built to solve.
What is sudiviz ?
sudiviz is a free, open-source Python CLI that not only visualises live cloud infra but also diagnoses & fixes them, giving you healthy status saving you money & time.
pip install 'sudiviz[all]' to your AWS account, and kaboom, your whole infra is in front of you without going through network spaghetti through AWS console. It renders the whole thing as a live topology graph using NetworkX and Cytoscape.js.Features :
Everything you need to actually see and fix your cloud.
Live Topology
Real-time graph: ALB → Target Groups → EC2 → SGs
Traffic Animation
Watch requests flow from ALB to backends live.
Health Heatmap
Spot unhealthy targets and failing services instantly
Cost Heatmap
Estimated monthly spend per resource on the topology
Orphan Detection
Find unattached target groups, unused SGs, dead ENIs
Terraform Drift
Compare live AWS vs terraform show -json output
Auto-Fix
sudiviz fix --apply remediates in one command
MCP / AI Agent
Plug into Claude or any MCP-compatible agent

The demo walked through four views:
1. Multi-resource topology. The full spaghetti, but drawn properly. You can see the load balancer, the target groups behind it, the instances, the security groups linking them, and how traffic is actually allowed to flow.

2. Multi-AZ layout. How resources are distributed, where your redundancy actually lives, and how traffic crosses zone boundaries. It's the kind of picture you never quite get from clicking through the Console one subnet at a time.

3. Cost heatmap. Every resource coloured by estimated monthly spend. Orphan resources – the target group nobody uses, the ENI attached to nothing – light up in a way that PDF invoices from Billing never quite manage.
4. Health view. The same topology, but coloured by health status. Unhealthy targets, missing ingress rules, RDS publicly accessible, storage not encrypted — all flagged by severity right on the graph.
Also a brief comparison between sudiviz vs leading market providers .

For the terminal people in the room
I asked who lives in the terminal all day. A lot of hands. So I switched over to
sudiviz tui , which is the Textual-powered terminal UI. Same infrastructure data, same diagnostic engine, but rendered in your shell — no browser, no context switch. If you're the kind of person who has tmux memorised and never willingly opens the AWS console, this is the mode for you.
The MCP angle
One thing I wanted the audience to walk away with: sudiviz has MCP support. Which means you can plug it into any MCP-compatible client and start asking your infrastructure questions in plain English. "What's unhealthy in production?" "Which security groups have no rules attached?" "Show me anything costing more than $50/month that hasn't been touched in a week."
The graph is still there. The CLI is still there. MCP just means the data is portable — you can fit it into whatever workflow you already use.
The live demo
I'll be honest — live demos scare me. There's a reason most conference talks lean on pre-recorded gifs. Anything can go wrong. The Wi-Fi. Your AWS credentials. The one region that decides to have a bad day at exactly the wrong moment.
But live demos are also the only way to prove the tool actually does what you say it does. So I pointed sudiviz at a real (deliberately messy) test account and ran through it in front of the room. Discover → topology → diagnose → fix.
It went well. Which I'm still slightly amazed by.
The questions that came after
Two questions stuck with me.
The first was about networking — specifically, how sudiviz figures out the security group flow between resources.
The second was one I hear a lot: can it do multiple AWS accounts at the same time?
No — sudiviz operates on one AWS account at a time. It uses a single
--profile / --region per invocation and resolves credentials through the standard boto3 chain (env vars, ~/.aws/credentials , SSO, instance profile). Although I will work on building this feature : DIf you need to cover multiple accounts, you'd run sudiviz separately for each one by switching the
--profile flag (e.g. sudiviz diagnose --profile account-a , then sudiviz diagnose --profile account-b ). You could also use the --json output to merge results from multiple runs yourself.Where to find it
- Website: pydevsg.github.io/sudiviz
- GitHub: github.com/pydevsg/sudiviz ( if you have come this far, please give sudiviz a star : D )
- PyPI:
pip install 'sudiviz[all]'
It's open source, GPL-3.0, and pure Python. If you want to contribute, the issues are open and the roadmap includes GCP and Azure support — because the visibility problem doesn't stop at AWS.
Huge thanks to the AWS London WAUG organisers for having me, to James Harding for organising this and giving me the opportunity and platform, and to everyone who came up afterwards with questions, feedback, and the occasional "I'm going to try this on my account tonight." That's the best thing you can hear as a maintainer.

#AWS #WellArchitected #OpenSource #Python #CloudInfrastructure #DevOps #SRE #Terraform #AWSCommunity #WAUG #sudiviz #CloudVisibility #FinOps #MCP
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article