Open Source

Your AI agent finally has a memory that works

graymatter gives your AI agents perfect recall without embeddings, GPU, or cloud dependencies. FTS5 indexing + LYT knowledge mapping = 70-98% token reduction on every query.

Install Now How It Works

Why graymatter

Built for agents that need to remember everything

Most agent memory systems rely on expensive embeddings and vector databases. graymatter uses battle-tested FTS5 full-text search + LYT (Linking Your Thinking) methodology.

🔍

Depth-Aware Search

Results ranked by recency + structural relevance. Know exactly how stale every result is before you use it.

🧠

LYT Knowledge Mapping

Maps of Content (MOCs) auto-generated from your notes. AI-powered organization that scales with your knowledge.

🌙

Nightly Reflection

Automatic synthesis at 2 AM. Yesterday's raw notes become distilled wisdom by morning.

⚡

Zero Cloud Dependency

SQLite + FTS5 runs anywhere. No API calls, no rate limits, no egress fees. Your data stays yours.

🪶

Context Saver Integration

Wraps any skill command. Returns compact summaries instead of flooding context. 70-98% token savings.

🔧

OpenClaw Native

Built for OpenClaw agents. Drop-in skill with CLI, API, and cron integration. Works day one.

Installation

Up and running in 60 seconds

graymatter is a skill for OpenClaw. Clone, configure, done.

# Clone the skill
git clone https://github.com/aldow3n-a11y/graymatter.git \
  ~/.openclaw/workspace/SKILLS/graymatter

# Configure your Obsidian vault path
echo "OBSIDIAN_VAULT=/path/to/your/vault" >> ~/.openclaw/.env

# Run initial sync
python3 SKILLS/graymatter/scripts/gm.py sync
            

Requirements: Python 3.10+, OpenClaw 2026.3+, Obsidian vault (optional but recommended)