Guardrails, rollbacks, and the craft of operating real apps on a cluster you own. Written for developers moving fast with AI who do not want to break things.
A real cluster does not cost 60 to 100 dollars a month. A single small VPS running k3s is a genuine cluster, and Burrow runs on it the same way it runs on a big managed one.
A real cluster does not cost 60 to 100 dollars a month. A single small VPS running k3s is a genuine cluster, and Burrow runs on it the same way it runs on a big managed one.
You have outgrown your PaaS on cost and lock in, and moving to a cluster you own sounds like a weekend of pain. Here is the real path, ten steps, honest about the work.
If you already run Loki or Prometheus, you do not have to install a second stack next to it. Burrow can connect to what you have and query it in plain language.
A cache stores the results of slow or repeated work so you can serve them fast. Here is when it helps, when it does not, and the full workflow to add one to your app.
Autoscaling should add copies of your app when traffic rises and remove them when it falls, between a floor so you are never caught at zero and a ceiling so a spike can never scale you into a surprise bill.
The clearest way to understand Kubernetes is to deploy one small web app on it by hand. Here is every piece you touch, explained plainly, and where the weight really lands for a solo developer.
Loading an MCP server into your agent's session puts its whole tool list in context, every turn. For an operations surface with many verbs, a CLI the agent runs on demand fits better. Here is the honest case.
When an agent operates your infrastructure, you should never have to take its word for what happened. The audit log is the record that lets you check.
Every app needs the occasional one off command: a migration, a seed, a cleanup script. Here is how to run one in your app's own environment, cleanly.
Take a deployed app from running only inside the cluster to live at your own domain over HTTPS, and diagnose it link by link when something is not ready.
How to set, update, and manage your app's configuration and sensitive values, with a clean line between the two and no restarts you did not ask for.
How Burrow lets an agent operate your live apps while it holds no cluster credentials, never carries your code, and cannot reach past the lines you draw.
A PaaS runs your app on their platform. Self hosting runs it on yours. Eight honest questions to figure out which one you actually want right now.
What it actually feels like to run a live app by talking to your agent: deploy, diagnose, fix, and keep control of the moves that matter.
Keep a loose place to move fast and a locked down place that stays safe, with the same commands across both. A practical guide to environments.
See what your app is doing and how it is performing, without assembling a full observability platform first. A guide to logs and metrics in Burrow.
Give your app a real database, wired in for you, plus a straightforward way to take and restore backups. A practical guide to the Postgres addon.
Burrow deploys your app by naming an image, not by taking your code. Here is what that means in practice and why it makes deploys safer and clearer.
The unglamorous operational work that stands between your code and a running app, and how Burrow takes each piece off your plate.
A start to finish walkthrough: install Burrow, deploy your app by image reference, check that it is healthy, and put it on the internet over HTTPS.
Approvals, denials, and warm rollbacks: the mechanisms that let Burrow hand an AI agent the keys to prod, safely.
Every deploy keeps the previous revision warm for 24 hours. Roll back with a single command, no rebuild, no cold start.
Our stance on data, telemetry, and lock in, and the choices that make self hosted mean what it says.
Sign off gates that batch, remember your intent, and never block a deploy that stays inside its limits.
Keeping the previous revision warm turns a bad deploy into a fast fix. Here is how it works, with an illustrative walk through.
The terminal is where developers already live. Every guardrail, approval, and rollback is one legible command away.
