By swen128
Persistent memory and scheduled autonomous tasks for Claude Code — memory files, cron scheduling, and workspace personality via MCP channels
npx claudepluginhub swen128/markus --plugin markusFirst-run onboarding. Use when SOUL.md does not exist in the workspace, indicating no prior markus setup. The SessionStart hook will nudge you to run this.
Summarize and deduplicate MEMORY.md to keep it manageable. Use when MEMORY.md grows large or contains redundant entries.
Search memory files for relevant prior context. Use before answering questions about past decisions, preferences, prior work, or anything previously recorded.
Extract noteworthy items from the current conversation into memory files. Use when a cron notification prompts periodic reflection, or when a decision, preference, or important context is observed.
Modify SOUL.md behavioral guidelines. Use when the user expresses a behavioral, tone, or style preference — including phrases like "from now on", "always", "never", "whenever you respond", "each time". Also use when a consistent behavioral preference is observed across multiple interactions. These are personality and response-style changes, NOT config/hook/permission changes — do not confuse with update-config.
Update USER.md with newly learned user information. Use when the user mentions their name, timezone, pronouns, preferences, or other personal details.
A Claude Code plugin that turns Claude into your personal assistant. It knows you, remembers your context across sessions, and acts autonomously on your behalf.
Requires Bun and a Claude.ai account (not API key).
/plugin marketplace add swen128/markus
/plugin install markus@markus
claude --agent markus:personal-assistant --dangerously-load-development-channels plugin:markus@markus
On first session, Markus introduces itself and asks about you — your name, preferences, how you like to work. Everything it learns carries over to the next session.
Markus works best when paired with a messaging channel so it can reach you outside the terminal. Install the Telegram or Discord plugin and launch with both:
claude --agent markus:personal-assistant --channels plugin:telegram@claude-plugins-official \
--dangerously-load-development-channels plugin:markus@markus
It remembers you. Share something once — your timezone, your coding style, a project decision — and Markus writes it down. Next session, it already knows.
You: My timezone is Asia/Tokyo and I prefer tabs over spaces.
(Markus updates your profile and remembers for next time)
--- next session ---
You: What's my timezone?
→ Asia/Tokyo
It reflects on what happened. Ask it to periodically review the conversation and save anything worth remembering — decisions, preferences, context.
You: Every 30 minutes, reflect on what we've been doing and save anything important.
(Markus schedules a cron task and reflects autonomously.)
It acts on a schedule. Cron tasks fire as channel notifications between your messages, so Markus can act without you prompting it.
You: Every morning at 9am, check my deployment status and summarize it.
(Markus schedules the task and runs it autonomously.)
It has personality. Markus maintains a SOUL.md that defines how it behaves — tone, style, values. You can shape it over time.
You: From now on, always respond in bullet points.
(Markus updates SOUL.md. Next session, it still uses bullet points.)
Persistent memory system for Claude Code - seamlessly preserve context across sessions
Admin access level
Server config contains admin-level keywords
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Qiushi Skill: methodology skills for AI agents guided by seeking truth from facts, with Claude Code, Cursor, OpenClaw, Codex, OpenCode, and Hermes guidance.
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Search and retrieve documents from local markdown files.
Intelligent prompt optimization using skill-based architecture. Enriches vague prompts with research-based clarifying questions before Claude Code executes them
Uses power tools
Uses Bash, Write, or Edit tools