ccmemo

Claude Code starts every session with a blank slate. The quirk you debugged yesterday, the plan you were halfway through — all gone when the session ends. ccmemo saves that knowledge and progress as Markdown files in your repository, so the next session picks up where you left off.

What It Does

ccmemo adds four slash commands that persist knowledge and plans as plain Markdown in .claude/:

Claude Code invokes these automatically when relevant; you can also trigger any of them by name.

Quick start

Up and running in three steps. The plugin is the recommended path — skills and hooks activate the moment it's installed.

1. Install the plugin (in Claude Code):

/plugin marketplace add LevNas/claudecode-plugins
/plugin install ccmemo@levnas-plugins

That alone makes /record-knowledge, /plan-task, /review-knowledge, and /recall-knowledge available and wires up the Context Guard hooks.

2. Scaffold the starter config into your project — a tag registry for knowledge and an index for tasks. Easiest is to just ask Claude Code:

Scaffold ccmemo's knowledge and tasks templates into .claude/.

Prefer a command? From your project root:

tpl=$(find ~/.claude/plugins/cache -type d -path '*ccmemo*/templates' | sort | tail -1)
cp -r "$tpl/knowledge" .claude/knowledge
cp -r "$tpl/tasks"     .claude/tasks

3. Try it. Run /record-knowledge and describe something worth remembering — Claude writes a Markdown entry under .claude/knowledge/entries/. Commit it, and your next session (or a teammate's) finds it automatically.

Semantic search via /recall-knowledge is opt-in (one-time index build) — see docs/hybrid-search.md.

Without the marketplace

Copy the skills and templates straight from the repo:

git clone --depth 1 https://github.com/LevNas/ccmemo /tmp/ccmemo
cp -r /tmp/ccmemo/skills/*            .claude/skills/      # the four skills
cp -r /tmp/ccmemo/templates/knowledge .claude/knowledge
cp -r /tmp/ccmemo/templates/tasks     .claude/tasks
rm -rf /tmp/ccmemo

Your project ends up with:

your-project/.claude/
├── skills/{record-knowledge,plan-task,review-knowledge,recall-knowledge}/SKILL.md
├── knowledge/{CLAUDE.md,entries/}
└── tasks/{CLAUDE.md,readme.md}

Hooks (Context Guard) are wired through the plugin; with a manual copy, add the hooks/hooks.json entries yourself if you want them.

Example

/record-knowledge creates an entry like .claude/knowledge/entries/20260302-143052-alice-docker-compose-port-conflict.md:

---
title: Docker Compose port conflict with host nginx
author: "@alice"
created: 2026-03-02
status: active
tags: "#docker #nginx #pitfall"
---

When running `docker compose up`, port 80 conflicts with the host's nginx.
Use `ports: ["8080:80"]` or stop host nginx first.

- ref: [Docker Compose networking](https://docs.docker.com/compose/networking/)
- see: [Nginx reverse proxy setup](nginx-reverse-proxy.md) — related configuration

See docs/examples.md for personal and team workflow walkthroughs.

Usage

Searching entries

Quick keyword/filename lookups need no setup:

fd -e md . .claude/knowledge/entries/ | fzf   # fuzzy search by filename
rg '#pitfall' .claude/knowledge/entries/      # search by tag
rg '^title:' .claude/knowledge/entries/       # list all titles

For meaning-based search (synonyms, or a Japanese query against English identifiers), /recall-knowledge runs hybrid semantic search — ripgrep + local vector embeddings + the see:-link graph — and falls back to ripgrep when the index is absent. Setup: docs/hybrid-search.md.

Reviewing knowledge

/review-knowledge keeps the base healthy in three modes:

• Health check (default) — reports stale entries (>90 days), orphans (no see:

links), missing connections, tag issues, and summary stats

• Topic review (topic:<keyword>) — summarizes a topic and asks reflective

questions to verify accuracy

• Fix mode (fix) — interactively adds missing links and registers tags

Plan & task management

/plan-task persists multi-step plans in .claude/tasks/. Two modes:

• Git-tracked (default) — plans (plan-v1.md, todo.md, readme.md) are committed

and become the shared source of truth. Progress moves [ ] → [~] → [x]; plan revisions are kept as plan-v2.md, etc. Each task dir also holds context-*.md files that capture detailed working context (see Context Guard).

• Issue-centric — gitignore .claude/tasks/ and treat your issue tracker (GitHub,

GitLab, Jira) as the source of truth; .claude/tasks/ becomes a per-session scratchpad. Session start checks assigned issues instead of readme.md.

Wiring into CLAUDE.md

Add a lookup section to your project's CLAUDE.md so Claude checks relevant entries before starting work. Patterns and examples: docs/claude-md-examples.md.

Customization

• Tags — maintain a registry in .claude/knowledge/CLAUDE.md (lowercase

kebab-case, # prefix). Claude checks it before creating new tags.

• Author — the entry author field defaults to @<username>; set it to your Git

hosting username.

• Workflow — edit any skills/*/SKILL.md to add project-specific conventions

(tag categories, issue-tracker comment formats, plan templates).

• Auto-commit (opt-in, off by default) — set CCMEMO_AUTOCOMMIT=1 and a

safety-net hook commits only .claude/knowledge/ and .claude/tasks/ changes at session end (SessionEnd) and before compaction (PreCompact). It never runs git add -A, never pushes, and a leak-scan gate blocks commits containing leak-prone shapes (set CCMEMO_AUTOCOMMIT_ON_LEAK=warn to commit with a warning instead). It complements — does not replace — your manual end-of-session commit, so it is a no-op when you have already committed.

Why Use It in a Git Repository

Everything is plain Markdown in your repo, so it follows the same branch/merge/review workflow as your code:

• Team sharing — a pitfall one person finds on Monday is available to everyone

(and every Claude Code session) on Tuesday.

• Session continuity — plans and knowledge survive across sessions; no

re-explaining context or re-discovering the same issues.

• Browsable — files render cleanly in GitHub and GitLab with no special tooling.

Documentation

• docs/architecture.md — subagent delegation, Context Guard, internals
• docs/examples.md — personal & team workflow walkthroughs
• docs/hybrid-search.md — semantic search setup & verification (incl. corporate TLS)
• docs/claude-md-examples.md — CLAUDE.md configuration

Getting help

• Usage or design questions — open a Question issue
• Bugs & feature requests — pick a template from new issue
• What changed — see the CHANGELOG

Contributing

Maintained by @LevNas; contributions and forks welcome. See CONTRIBUTING.md for scope, plugin conventions, the linter command, and pull request guidelines.

License

MIT. If you build on ccmemo, a credit line is appreciated (not required):

Based on ccmemo by LevNas.