DevOps diagnosis you can talk to
What’s broken? Ask Trace.
Describe the problem. Trace connects to GitHub, your cloud and your GPUs, finds the cause, and fixes it once you approve.
- GitHub
- Google Cloud Run
- AWS
- RunPod
- Lightning AI
How it works
Three steps. You stay in charge the whole way.
Step 1: Ask Trace
Describe the problem like you would to a teammate: “checkout is failing”, “my GPU job crashes”.
Step 2: Connect only what it needs
Trace asks for just the accounts that matter for your question, read-only. Then it checks deploys, settings, logs and the live site, and shows every step.
Step 3: You approve the fix
Trace explains the cause in plain English and gets a fix ready. Nothing changes until you say OK.
See Trace work
A real question, answered in plain English. Every step is shown, and the fix waits for you.
What it catches
The problems that quietly eat an afternoon.
Staging works, production doesn’t
Settings that quietly differ between environments: a missing variable, an old secret, a different region.
For example: DATABASE_URL set in staging, missing in production
“Blocked by the browser”
CORS errors after a deploy, traced back to the exact setting and the change that caused them.
For example: shop.example.com not in ALLOWED_ORIGINS since yesterday’s deploy
GPU jobs that won’t start
CUDA and driver versions that don’t match the image you shipped, on RunPod or Lightning AI.
For example: Image needs CUDA 12.4 · 2 of 4 servers have driver 535
Failures that jump between services
The error shows up in one place but starts in another. Trace follows it across services and providers.
For example: checkout → payments → orders database
Safe by default
Trace is built to look first and ask before it touches anything.
Only looks by default
Every connection starts read-only. Trace reads, it doesn’t write.
Nothing changes without your OK
Fixes wait for someone on your team with permission to approve them.
Secrets stay hidden
Values are hidden before Trace reads anything. It sees setting names, not passwords.
Every action is logged
Each command Trace runs is shown with its output and kept in your activity log.
You choose what it can see
Connect only the accounts and services you want. Disconnect any of them at any time.
Works where you work
Web app, on any device
Chat with Trace and approve fixes from your laptop or your phone. Add it to your home screen to get a nudge when something needs you.
VS Code and Cursor
Ask Trace about the code you have open, without leaving your editor. Sign in once with a token from the web app.
Questions, answered
What does Trace actually do?
It investigates deploy and runtime problems across your code and cloud accounts, tells you the cause in plain English, and gets a fix ready for you to approve.
Can Trace change things on its own?
No. It only looks by default. Any change, like a setting, a restart or a pull request, waits for someone on your team to approve it.
Which providers does it work with?
GitHub, Google Cloud Run, AWS, RunPod and Lightning AI. Connect one or all of them; Trace asks only for what it needs to answer your question.
What can Trace see?
Only what you connect. Secret values are hidden before Trace reads anything, and every command it runs is logged so your team can see it.
Do I need to install anything?
No. Trace runs in your browser on any device. The VS Code and Cursor extension is optional, for asking about the code you have open.
How much does it cost?
Trace is free while it’s in early access. We’ll give plenty of notice before any paid plans.
Next time a deploy breaks, ask Trace.
Connect read-only in a few minutes. Nothing changes without your OK.