I kept losing context between Claude Code sessions. I was building a SaaS app — 50+ sessions deep — and every new conversation started cold. Claude didn't remember what we built yesterday, what architecture decisions we made, or what security issues we'd already caught. I was re-explaining the project every time, wasting tokens and repeating mistakes we'd already solved. The real frustration was that all the knowledge existed — in git commits, in code patterns, in decisions scattered across sessions — but it evaporated the moment a session ended. So I built Shiplog: a Claude Code plugin that gives every project persistent memory. It watches your commits and silently runs background agents that log architecture decisions, auto-discover features, review security, and track code patterns. When you start a new session, Claude already knows where you left off. The approach evolved significantly. I started with a "planned mode" where you'd define a full PRD upfront. But I realized most solo devs just want to build — they don't want to write specs before writing code. So I added "vibe mode" as the default: just code, commit, and Shiplog figures out what you're building from the diffs. No planning required, no workflow friction. The agents do the bookkeeping you'd never do yourself. The core philosophy stayed simple throughout: your project's history shouldn't disappear when a session ends.

Shiplog

Stop re-explaining your project to Claude

Stop re-explaining your project to Claude