🧠 Second Brain

An agent that actually remembers.

Every AI assistant has amnesia. You tell it your name, your stack, your preferences — and next session, it's gone. You start over. Again.

Second Brain fixes this. It's a drop-in memory system for OpenClaw that makes your agent remember what matters — preferences, project context, decisions, and patterns — across every session, automatically.

No databases. No embeddings. No infrastructure. Just well-structured markdown files that your agent reads and writes as part of its normal workflow.

What Changes After Install

How It Works

Three simple rules drive the whole system:

1. Preferences are captured immediately. Say "I prefer Tailwind" once, and it's written to memory/preferences.md right then. You'll never be asked again.

2. Sessions are summarized automatically. After meaningful work, the agent appends a concise summary to today's daily note (memory/YYYY-MM-DD.md). Decisions, outcomes, and follow-ups — captured.

3. Weekly consolidation keeps it clean. Every week, daily notes are distilled into MEMORY.md and archived. Your agent's memory stays fast and relevant, not bloated.

Memory File Structure

your-workspace/
├── SOUL.md                 # Personality: natural recall, no redundant questions
├── AGENTS.md               # Memory workflow: when to capture, consolidate, prune
├── TOOLS.md                # File structure conventions and formats
├── MEMORY.md               # Consolidated long-term memory (auto-maintained)
└── memory/
    ├── preferences.md      # Your preferences, grouped by category
    ├── projects/           # Per-project context files
    │   └── my-app.md
    ├── archive/            # Processed daily notes
    └── 2024-01-15.md       # Today's session notes

Copy SOUL.md, AGENTS.md, and TOOLS.md to your OpenClaw workspace root. The memory/ folder is created automatically on first use.