Hey Product Hunt! I'm the creator of auto-sop.

I built this because I was frustrated with Claude Code making the same mistakes over and over. I'd fix a bug, and
two sessions later Claude would introduce the same issue. Same wrong import path. Same missing error handler. Same forgotten validation.

auto-sop fixes this with a self-improving feedback loop:

1. It watches your Claude Code sessions (locally, nothing leaves your machine)
2. Every 10 turns, it analyzes patterns and detects recurring mistakes
3. It auto-generates CLAUDE.md directives that Claude follows in future sessions
4. Critical errors graduate instantly — one observation is enough

The result? Claude Code that genuinely learns from its history. We're seeing 30+ directives per project and
measurable token savings from fewer retry loops.

The core CLI is free and source-available under the Elastic License 2.0. Pro ($12/mo) adds unlimited projects,
cloud sync, and analytics dashboard.

Some things I'm proud of in the architecture:

- Zero latency impact on your Claude sessions (async hooks, fail-open)
- Privacy-first: raw captures never leave your machine
- Works with your existing Claude subscription — no separate API key needed
- Event-driven learning: no cron job hammering your CPU, fires only when there's enough data

Shout-out to @anthropic and the Claude Code team — auto-sop is built with Claude Code, for Claude Code. The irony
isn't lost on me: an AI built the tool that teaches itself not to repeat mistakes. 🦉

I'd love to hear your feedback! What mistakes does Claude Code keep making for you?

auto-sop

Your Claude Code learns from its mistakes automatically

Your Claude Code learns from its mistakes automatically