Skip to main content

Getting started

From alert to an evidence-backed decision

Production Master is a configurable AI-SRE pipeline you can run anywhere: it turns approved production signals into a cited, replayable evidence dossier and leaves consequential actions behind a human approval gate.

Start hereproblemevidence modelplugin repo

Why this exists

Incident response gets expensive when context disappears.

Production Master is built for teams that need to understand what happened without turning every investigation into a senior engineer’s private reconstruction exercise.

The problem

The incident is not in one system

An alert tells you something is wrong.
The explanation is distributed across logs, metrics, deploy diffs, tickets and the context held by whoever happened to be on call.

The shift

Make the reasoning inspectable

Production Master gathers those approved signals into exhibits, forms falsifiable hypotheses, and records what each path proves or rules out.

The decision

Stop before the consequential action

The result is a cited, replayable dossier and a proposed next step.
A human reviews, corrects, branches or approves it before anything changes in production.

What you get

The selling point is not more automation.

It is better operating memory: a pipeline that keeps the evidence, the uncertainty and the approval boundary visible after the incident has moved on.

Pipeline as data

Your investigation playbook, made explicit

Triggers, evidence windows, tools, budgets and hand-offs are configuration — not a prompt hidden in a chat transcript.

Evidence and replay

A conclusion that can be read back

Claims retain source pointers, timestamps and the rejected paths that shaped the final explanation.

Human gates

Useful automation without silent side effects

Production Master prepares a recommendation, pauses at the boundary, and lets an accountable engineer approve, correct or branch the investigation.

Run anywhere

Keep the operating model yours

Use the managed service, self-host via Helm, bring your own model keys, and start with the public IDE plugin surface.

Start with the public surface

A clear path from curiosity to one real run.

The product is designed to meet engineers where they work. Use the web experience for the investigation and the plugin surface to carry the same workflow into the IDE.

Read the product story

Start with the evidence model and the approval boundary, then bring the workflow to one incident your team already understands.

Explore the product

Open the plugin repo

The public repo is the starting point for the IDE surface across Claude Code, Codex, Cursor and OpenCode.
Platform-specific setup lives in its README.

Open production-master

Bring a real incident

Closed-beta teams start read-only, define the evidence window and review the first reports critically.
The pilot is designed around your operating boundary.

See the closed beta

Claude Code marketplace

The plugin README keeps the install paths current across editors. The marketplace command is:

/plugin install production-master@production-master-marketplace --scope project

Make the next incident legible

Bring the evidence gap, the operating boundary and the engineer who will challenge the first answer.

Read the public plugin repo first; closed-beta access starts with one real incident.