A DevOps engineer that lives on your machine
Type English.
Deploy to your own cloud.
PlainOps connects to your AWS, Google Cloud, or Azure account and your own Anthropic API key. Describe the app, read the monthly cost, click Approve — get a verified live URL. Then keep it alive: diagnosis, rollbacks, migrations, backups, monitoring.
Free to run · you bring your own cloud + AI keys · no account, no server, no telemetry
The promise is the security model
Not a SaaS. There is no PlainOps server. Everything runs on localhost — auditable in the open source.
Credentials stay home
Deploys use your local aws / gcloud / az login. Nothing is uploaded anywhere except calls to your own cloud and your own Anthropic key.
The AI never sees a secret
Secret values live in an AES-256-GCM vault on your disk and go straight to your cloud's secret store. The model only ever sees {{secret:NAME}} — every string is scrubbed first.
You approve every change
The AI proposes; you click; the machine executes. Every mutating action waits for a real click — approval is a code path, not a prompt rule. Reads are instant, writes wait for you.
What a DevOps engineer does, it does
Every shape of app
Monoliths with a database, serverless APIs with queues, many-service microservices, static sites — real infrastructure from reviewed OpenTofu blueprints, never AI-authored Terraform.
Cost before anything exists
A monthly estimate on the approval card before a single resource is created — plus a live cost dashboard with real AWS spend.
Plain-English diagnosis
"Something's broken" collects the live URL probe, service state, real logs, and recent actions — then explains the root cause and the fix like a colleague would.
Day-2, not just day-1
Safe deploys with automatic rollback, linted snapshot-first DB migrations, one-command rollback, secret rotation, cron jobs, backups + restore drills, drift detection, cloud-resident monitoring.
Guardrailed CLI access
Ask anything about your infra — read-only commands run instantly, mutating ones need your click, credential-exposing ones are refused. On all three clouds.
Adopts what already runs
Point it at infrastructure it never deployed and it sweeps the whole region or project: unhealthy services, error logs, firing alarms, waste.
Three steps to live
Describe
"Deploy C:\code\my-store with a Postgres database" — or paste an architecture diagram and it builds the app too.
Approve the cost
Read the plan and the monthly estimate, then click. Nothing is created before this moment.
Get a verified URL
No deploy is called live until the URL answers a real HTTP 200 — measured, never hoped.
Download PlainOps
Grab the installer for your machine — the app is the full product, nothing held back.
macOS
detectedmacOS 11+. Unsigned build for now — first open: right-click the app → Open.
Linux — or run anything from source
git clone https://github.com/shobhit9957/PlainOps.git
cd PlainOps && npm install
npm start # dashboard at http://localhost:7717
npm run demo # scripted walkthrough — no cloud or API key neededNode 20+. Full guides in the repo: README, QUICKSTART, per-cloud CLI setup, architecture, security model, troubleshooting. All releases →
Quick answers
What does it cost?
PlainOps itself is free to download and run, source-available on GitHub. You pay your own cloud bill and your own Anthropic API usage — both stay in accounts you control, with the cost shown before every deploy.
What do I need?
An AWS / Google Cloud / Azure account with its CLI signed in, and an Anthropic API key. The app checks your setup and walks you through what's missing.
Does it phone home?
No. There is no PlainOps server, no account, no telemetry. The only network calls are to your own cloud and the Anthropic API with your key.
What if the AI is wrong?
It can't act alone: infrastructure comes from reviewed blueprints, every change waits for your click, and "live" is a measured HTTP 200 — not an optimistic message.