AI-powered CLI tool

Never write a post-mortem // git log → 3 questions → perfect document

It's 2am. You just fixed a production incident. The last thing you want to do is write a post-mortem. deathlog does it for you in 30 seconds.

30s
to generate
3
questions asked
0$
open source
hours saved
live demo

One command.
Complete post-mortem.

Run deathlog inside any git repo. That's it. Claude reads your history and builds the document.

~/my-startup — zsh
deathlog
💀 deathlog v1.0.0 — AI Post-Mortem Generator
Scanning git history...  found 63 commits (last 30 days)
Reading error logs...   found 2 critical, 5 warnings
Last deployment:     3h 42m ago → v3.1.2
Recent changes:     payment-service, auth, db migrations
? What broke? Stripe webhooks stopped processing
? Duration + users affected? ~90 min, ~340 users, $2.1k revenue impact
? Is it resolved? yes — rolled back payment-service to v3.1.1
Analyzing with Claude AI...
Timeline reconstructed from git log
Root cause identified: breaking change in stripe-sdk v5.2.0
4 action items generated
Post-mortem saved → postmortem-2026-03-06.md
Done in 4.1s  |  Go sleep. You earned it.
how it works

Reads your repo.
Writes the doc.

deathlog connects your git history to Claude AI and generates a structured, professional post-mortem in seconds.

01
🔍

Reads Context

Automatically scans your git log, recent commits, deployment history, and error logs to understand what happened.

02
💬

Asks 3 Questions

What broke? Who was affected? Is it resolved? That's all we need. No forms, no tickets, no long surveys.

03
🤖

Claude Generates

Sends everything to Claude AI with a battle-tested prompt. Gets back a structured, professional post-mortem document.

04
📄

Saves as Markdown

Outputs a clean .md file ready to paste into Notion, Confluence, Linear, or email to your team. Done.

before vs after

The difference is real.

Here's what writing a post-mortem looks like before and after deathlog.

❌  Without deathlog
😩

Spend 2 hours digging through git log trying to reconstruct the timeline

🔍

Manually search Slack for what people said during the incident

📝

Write it all up from scratch in a blank document

🥱

It's now 4am. You're exhausted. The doc is terrible.

🚫

Most teams just skip it entirely and repeat the same mistakes

✅  With deathlog

Run one command from your terminal — takes 4 seconds

🤖

AI reads your git history and reconstructs the timeline automatically

💬

Answer 3 quick questions — 30 seconds total

📄

Get a complete, professional Markdown document instantly

😴

Go sleep. The post-mortem is done. Your team learns from it.

features

Built for engineers
who ship fast.

Everything you need. Nothing you don't.

🔄

Git-aware context

Automatically reads your last 30 days of commits, branches, and tags to understand your deployment history.

Instant output

From command to complete document in under 10 seconds. No loading screens, no dashboards, no accounts.

📋

Structured format

Always outputs: Summary, Timeline, Root Cause, Impact, Action Items. Consistent every single time.

🔌

Works everywhere

Any git repo, any stack, any language. macOS, Linux, WSL. No config required — just install and run.

🔓

Open source

100% open source on GitHub. MIT license. Read the code, fork it, improve it, make it yours.

🚀

Coming soon: Slack

Auto-post your post-mortem to a Slack channel. Tag team members. Create Linear issues. One command.

get started

Ship your first post-mortem
in under a minute.

No account. No dashboard. One command.

Step 1 — Install

pip install deathlog copy

Step 2 — Set your API key

export ANTHROPIC_API_KEY="sk-..." copy

Step 3 — Run it

cd your-project && deathlog copy

Requires Python 3.11+ and an Anthropic API key. ~$0.01 per post-mortem.