remember

Remember.md — One Brain. Every AI Tool.

Your AI tools have memory. But memory is not a brain.

Remember.md is a second brain plugin for OpenClaw and Claude Code. It organizes decisions, people, projects, and tasks from your AI sessions — past and future — into a structured, Obsidian-compatible knowledge base that travels with you across tools.

Free. Local. Open source. Portable.

---

Memory vs Brain

Every AI tool now has memory — flat notes you can't search, browse, or take with you. Remember builds something different: a structured second brain with people, projects, decisions, and tasks connected via wikilinks.

	Built-in memory	Remember.md
Structure	Flat key-value pairs	People, Projects, Notes, Tasks, Journal
Connections	None	[[wikilinks]] across all files
Browsable	No	Obsidian vault with graph view
Portable	Locked to one tool	One brain, every AI tool
Past sessions	No	Process months of history retroactively
Your patterns	No	Persona.md learns your code style

---

Install

OpenClaw

openclaw plugins install @remember-md/remember
/remember:init

Claude Code

Run these from your terminal (not from inside Claude Code):

claude plugin marketplace add remember-md/marketplace
claude plugin install remember@remember-md

Then start Claude Code and run /remember:init to create your second brain structure and configure permissions.

Important — use the terminal, not the /plugin slash command. An unrelated plugin also named remember exists in the default claude-plugins-official marketplace. The /plugin install remember@remember-md slash command currently ignores the @remember-md qualifier and installs the wrong one. The terminal command (claude plugin install remember@remember-md) honors the marketplace and installs correctly. If you already got the wrong one, remove it with claude plugin uninstall remember@claude-plugins-official.

---

What You Get

~/remember/
├── REMEMBER.md     # Your custom rules (you edit this)
├── Persona.md      # Your patterns + Top Beliefs (AI auto-manages)
├── People/         # One note per person (type: observation)
├── Projects/       # Active work with logs and tasks
├── Notes/          # World-facts, decisions, beliefs (typed)
├── Journal/        # Daily notes (YYYY-MM-DD.md)
├── Tasks/          # Focus + Next Up priorities
├── Areas/          # Ongoing responsibilities
├── Resources/      # Links, articles, references
├── Inbox/          # Quick capture
├── Templates/      # Note templates
└── Archive/        # Completed projects

All files use YAML frontmatter + [[wikilinks]] — Obsidian-native, browsable in any markdown editor.

Plugin state (audit log, thresholds) lives at ~/.local/state/remember/ — outside the brain, never touched by Obsidian.

---

Commands

Command	What it does
/remember:init	Initialize your second brain structure + auto-configure MCP search (run once)
remember this: ...	Instant capture — routes to the right place automatically
/remember:ask <question>	Ask the brain. AI reformulates as an optimal search query, calls the MCP search_brain tool, and answers with citations. Requires the @remember-md/mcp server (auto-configured by /remember:init).
/remember:status	Show brain stats — file counts, freshness, top beliefs
/remember:process	Optional. Extract knowledge from past AI sessions into your brain
/remember:evolve	Periodic (weekly recommended). LLM-driven brain evolution — entity profile synthesis, belief confidence re-scoring, stale detection. Schedule via /loop 7d /remember:evolve

---

How It Works

Process old sessions

Run /remember:process and recover months of lost knowledge from past OpenClaw and Claude Code sessions:

Found 47 unprocessed sessions.

✓ Extracted People/sarah-chen.md
✓ Extracted Notes/decision-database.md
✓ Created 12 journal entries
✓ Updated Tasks/tasks.md (+8 tasks)
✓ Updated Persona.md (learned your patterns)

Instant capture

Say "remember this: met with Sarah, decided to use Postgres for ACID compliance" and Remember routes it:

• Person → People/sarah.md
• Decision → Notes/decision-database.md
• Task → Tasks/tasks.md

Adaptive Persona

Persona.md evolves with you — code style, naming conventions, review preferences, communication patterns. Loaded automatically every OpenClaw and Claude Code session so your AI knows how you work.

Self-evolving brain

Captured facts aren't static. The brain has three layers and an epistemic schema (type / freshness / confidence / evidence) that lets it grow: