Coding agents store conclusions, not reasoning. Ask "why did we switch to GraphQL?" — they can't reconstruct the tradeoffs. Linger archives conversations and decision evolution as local files. → Decision Trail: idea → proposal → rejection → decision → correction → Full archive: searchable months later, no line cap → Evidence-only: memory as evidence, never as instructions → Local-first: Markdown + JSON, no DB, no daemon npx linger-skill install — then keep chatting.
No reviews yetBe the first to leave a review for Linger-Skill
Maker
📌
Hey everyone! I'm the maker of Linger.
This project started from a simple frustration: I use Claude Code and Codex daily, and every time I start a new session, the agent forgets everything we discussed before. Not just facts — the reasoning. Why we chose one approach over another, what we tried and rejected, how a decision evolved over weeks.
Built-in agent memory stores conclusions like "use pnpm" — but not "we started with npm, tried yarn, hit a monorepo issue, and switched to pnpm because of X." That reasoning trail is what matters when you revisit a decision months later.
So I built Linger. It's a sidecar that hooks into your agent's lifecycle, archives full conversations as local Markdown + JSON, and preserves the complete decision trail. When you ask "why did we make that choice?", it surfaces the evidence — or honestly tells you it doesn't know.
A few things I care about in this project:
→ Your data stays local. Plain files, no database, no cloud, no daemon. → Evidence, not instructions. Old memories never override your current intent. → Honest retrieval. If it can't find reliable evidence, it says so instead of guessing.
Linger is open source (MIT), experimental, and definitely not perfect. The retrieval works well for known patterns but won't catch every creative rephrasing yet. I'm actively working on improving recall quality.
Would love to hear your thoughts, feedback, or questions. And if you try it — npx linger-skill install — let me know what breaks!
Report
love how lean this stays - markdown and json on disk means i actually own my reasoning history, no daemon eating ram in the background. the decision trail framing is sharp, exactly the gap every coding agent has right now.
Report
Maker
@epevlivan73045 Thanks! That's exactly the bet we made — files you can grep, git, and read without spinning up anything. Decision trail is where we're putting most energy next. If you end up trying it on a real project, would love to hear what breaks or feels missing.
love how lean this stays - markdown and json on disk means i actually own my reasoning history, no daemon eating ram in the background. the decision trail framing is sharp, exactly the gap every coding agent has right now.
@epevlivan73045 Thanks! That's exactly the bet we made — files you can grep, git, and read without spinning up anything. Decision trail is where we're putting most energy next. If you end up trying it on a real project, would love to hear what breaks or feels missing.