---
title: "DevOps project management with Claude and a fleet · Projectify"
description: "DevOps project management for platform teams: toil, upgrades and postmortem actions in one queue, with a Claude Code fleet opening pull requests overnight."
url: https://projectify.si/for/devops
---

# DevOps project management for the work between incidents.

> Toil, upgrades, certificate renewals and postmortem actions sit in one ranked queue. Claude takes the next ready task, and on your own machine a fleet turns the small ones into pull requests overnight.

Start free: https://app.projectify.si/signup. A month of Premium free, no card. Agents never take a seat.

## Why the toil never gets done.

- "Toil never makes the sprint because incidents eat the week." Keep it ready in one queue. Fleet workers take the small chores overnight and open pull requests you review with your coffee.
- "Our runbooks are out of date the week after we write them." Record the fix as you do it. Projectify turns the recording into a workflow with steps, prompts and checks, and the runbook is read so anyone can ask it.
- "Postmortem action items vanish once the meeting ends." Paste the postmortem into chat. Each action becomes a task with an owner and the sentence it came from, ranked against everything else.

## An extra engineer for the chores, on your own machine.

- **Clears toil while you sleep.** Dependency bumps, pinned images, flaky tests and config tidying: fleet workers take them from the ready queue, each in its own worktree, and never merge. For example: HLY-55: pull request open, CI green, waiting in Review.
- **Answers from your runbooks.** Runbooks, architecture notes and past postmortems are read when they are added, so @Claude in chat answers with the page it found it on. For example: Failover runbook, step 4: promote the replica, then repoint the pooler.
- **Turns a postmortem into work.** Paste the timeline and notes into the project’s chat. Claude proposes the actions as tasks, and you keep the ones that matter. For example: Five actions from Tuesday’s 502s, two already ranked high.
- **Runs a production change the same way.** Plan, second pair of eyes, staging, checks, apply. A change workflow stops at each gate and waits in Approvals for a person. For example: Queue module rollout: staging done, production apply waits for Theo.

## The board it starts with

- Statuses: Triage, Ready, In progress, Review, Rolled out, Done
- Fields: Service, Environment, Severity, Change type, Pull request, Owner
- Views: Board, Table, Calendar, Tracker, Dashboard
- A workflow, Production change: Write the change plan → Second engineer reviews the plan (a person checks) → Apply to staging → Smoke tests and dashboards → Approve the production apply (a person checks) → Change log entry

## Before you give it the keys.

### Will an agent touch production?

Fleet workers only open pull requests from their own worktrees and never merge. Applying a change stays with your team and your pipeline.

### Can we keep the on-call checklist here?

Yes. A routine is a recurring checklist ticked by day, so each shift sees what the last one checked.

### Does it replace our incident or paging tool?

No. It holds the work around incidents: actions, toil, runbooks and upgrades. Keep paging where it is.

### Which code tools does it work with?

GitHub branches and pull requests show on their tasks, and Azure DevOps, Trello and Monday boards sync both ways.

Related: [Web development](https://projectify.si/for/web-development.md), [Engineering firms](https://projectify.si/for/engineering.md), [SaaS](https://projectify.si/for/saas.md), [SOPs](https://projectify.si/for/sops.md), [Sprint planning](https://projectify.si/for/sprint-planning.md). Every kind of work: [For your work](https://projectify.si/for.md).