Your Claude memory companion. Turn it on, keep it lean, schedule the upkeep, reverse anything.
Clean the auto-memory directory in a single pass: triage → show summary → apply (with backup). Saves the user's backup folder preference on first run so subsequent runs are one command.
Apply a pre-approved triage plan to a memory directory.
Distill a single memory file in-place — replace prose narrative with 1-3 sentences of forward-looking truth.
Triage an auto-memory directory as a **read-only dry run**. No file modifications.
Schedule a recurring `/mempenny:clean` for the current memory directory. The schedule lives in `~/.claude/mempenny.config.json`; the cleanup itself fires via a plugin-shipped `SessionStart` hook the next time you open Claude Code in this project after the scheduled time.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Your Claude memory companion. Turn it on, keep it lean, schedule the upkeep, reverse anything.
Claude's memory grows. Old notes pile up. The signal gets buried. MemPenny tidies it — and Claude's next session starts sharper.
Before: 424 files · 1,247 KB loading every session
After: 227 files · 458 KB · ~63% lighter
A real second-pass run; full case study: Real-world results.
Two ways:
/mempenny:clean. One command. Minute or two. You see the proposal, say yes, done./mempenny:nap. Pick a schedule (daily / weekly / once). MemPenny tidies on your next Claude Code session — backup-first, no prompts, fully reversible. Each pass leaves Claude with cleaner notes to learn from next time.What it does:
Don't like a change? /mempenny:restore puts everything back. Backup-first, always.
/plugin marketplace add marcelopaniza/mempenny
/plugin install mempenny@mempenny
/reload-plugins
/mempenny:clean — clean nowOne command, end-to-end:
/mempenny:clean
What happens, in order:
Yes, apply / No, cancel / Show full table. Nothing on disk changes until you say yes./mempenny:restore.First run in a memory directory asks where to keep backups (defaults to a sibling folder next to the memory dir) and remembers the choice. Every subsequent run is one command.
Roll back any time:
/mempenny:restore
Lists backups, you pick one. The current state is snapshotted first — the restore itself is reversible.
/mempenny:nap — schedule a learning napPick a frequency and a time. MemPenny runs /clean --yes for you when the schedule fires — no prompt, just a clean memory waiting for you.
/mempenny:nap # configure: backup folder → frequency → time
/mempenny:nap --list # show all configured schedules
/mempenny:nap --cancel # remove the schedule for this memory dir
Three questions, no daemons, works with whatever auth Claude Code already uses — OAuth or API key, both fine. The hook installs with the plugin and never modifies your ~/.claude/settings.json — MemPenny only writes that file when you explicitly accept the auto-memory enable offer.
When the schedule fires, MemPenny runs /mempenny:clean --yes on your next Claude Code session in this project — no Yes/No prompt, just a clean memory waiting for you. Backup-first, fully reversible via /mempenny:restore. Uses Claude credits per fire (same as a manual /clean).
Cross-platform: Linux + macOS for now. Windows support deferred.
/mempenny:clean--dir <path> — operate on a memory directory other than the current project's.--only <glob> — restrict triage scope by filename glob. Comma-separate multiple globs. Example: --only "project_*_20*.md,reference_*.md".--lang <code> — locale for user-visible output. Ships with en, es, pt-BR. Also honors MEMPENNY_LOCALE.--reconfigure — re-prompt for this memory directory's backup folder (ignores the saved entry). Other projects' entries are left alone.--yes — skip the confirmation gate; auto-apply after triage. Backup-first. This is what /mempenny:nap fires when the schedule runs./mempenny:memory-triage [--dir <path>] [--only <glob>] [--lang <code>] — dry-run triage. Produces a markdown classification table at a private mktemp path with permissions 600. No writes./mempenny:memory-apply <table-file> [--dir <path>] [--lang <code>] — applies a previously approved triage table. Table path is required; pass the path printed by /mempenny:memory-triage. Creates a backup before modifying anything./mempenny:memory-distill <file> [--lang <code>] — one-off distillation of a single file. Interactive: shows the proposal, asks to apply / skip / edit.Lock a whole folder — drop an empty .mempenny-lock file in it:
touch .mempenny-lock
/clean and /nap refuse to touch anything inside.
Lock a single memory file — add this line at the top:
<!-- mempenny-lock -->
npx claudepluginhub marcelopaniza/mempenny --plugin mempennyCompress Claude instruction files (CLAUDE.md, memory files, SKILL.md) into denser YAML. Verified lossless by meaning via round-trip review. Runs inside Claude Code; no API key required.
Toggle TL;DR response mode. When on, Claude answers maximally concisely unless you explicitly ask for a full explanation.
Lightweight cross-terminal messaging for Claude Code. Multiple sessions on one or many machines share a folder; messages arrive on the next prompt, at session start, and — on by default — the instant they land, waking an idle session in real time. Zero token cost when no traffic.
Lightweight cross-terminal messaging for Claude Code. Multiple sessions on one or many machines share a folder; messages are picked up by a hook on the next user prompt — zero idle polling, zero token cost when no traffic.
The highest-accuracy memory system for AI agents — 100% retrieval on LongMemEval. 14 content-aware summarizers, hybrid search (BM25 + vector + LLM judge), entity intelligence, decision trails. Fully local, zero cost.
Curated persistent memory for Claude Code. Write gate prevents bloat — only behavior-changing facts get saved. Tiered architecture: daily logs, structured registers, and auto-loaded working memory.
Persistent long-term memory for Claude Code via MCP — captures coding decisions, bugfixes, and context across sessions. Hybrid FTS5 + TF-IDF search with episode batching. Single SQLite DB, no external services. A lighter, lower-cost alternative to claude-mem (episode batching + a smaller model; cost savings are an internal estimate, not a measured benchmark).
Persistent memory across Claude Code sessions using Cognis
Persistent memory system for Claude Code with enforced context
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses