AI Memory Infrastructure

The memory layer
for AI agents.

graymatter gives your AI agents persistent, searchable memory. 70–98% token reduction, 19,500+ facts indexed, instant FTS5 recall — all running locally, no GPU required.

19,500+ Facts Stored
70–98% Token Cut
100% Local
FTS5 Instant Recall

Active · Production

graymatter

A high-performance semantic memory server that reduces context token usage by 70–98%. Built on FTS5 indexing with LYT mapping, it stores facts extracted from AI agent sessions and retrieves them in milliseconds. No GPU, no cloud, no API keys — just a local SQLite database and a Python server.

Engine
FTS5 + bge-m3
Storage
SQLite · AES-256-GCM
Embeddings
1024-dim · 100% coverage
Pipeline
Extraction worker · cron
Entities
19,192 tracked
Relations
27,344 mapped

The Knowledge Stack

The ecosystem around graymatter — tools, frameworks, and infrastructure for local-first AI PKM.

Core

CORTEX

Automated memory management with tier-based retention and Obsidian synchronization. Turns raw interaction logs into curated wisdom.

SQLite + Python · Tiered Synthesis

Enter the Cortex
Evolution

SecondCortex

Semantic triple extraction, vector search, and procedural memory. The next generation of agent memory — all free, all local.

Gemini Flash · Embedding-001 · SQLite

Explore SecondCortex →
Infrastructure

OpenClaw Deploy

One-command installer for a production-grade AI gateway. Your brain, always online and secure.

Alibaba Cloud · VPS · Tailscale-hardened

Deploy Now Guide
Framework

LYT + DodoMP

Personal knowledge management using the ACE framework and AI-generated Maps of Content.

Obsidian-native · AI Synthesis

Explore Methodology
Edge

Koda

The original 4WD companion robot. BLE control, computer vision, direct LLM connection. Predecessor to OttoBot v2.1.

ESP32-S3 · BLE · CV · LLM

Meet Koda
Tools

AI Coding Tools

Honest, battle-tested reviews of the best AI coding tools in 2026 — from Cursor to OpenCode.

Affiliate Recommendations

View Tools →
Stability

OpenClaw Guide

A practitioner's guide to maintaining high-availability AI gateways and hardening agent infrastructure.

Operational Excellence

Read Guide
Edge

OttoBot v2.1

ESP32-S3 companion robot with browser-ROS bridge and LLM brain. Audio streaming confirmed. MCP server with 10 tools.

ESP32-S3 · OV5640 · I2S · MCP

OttoBot v2.1 →

Active Projects

Current work across robotics, business, and content creation.

Active

🤖 OttoBot v2.1

ESP32-S3 companion robot with browser-ROS bridge and LLM brain via OpenClaw agent. Audio streaming breakthrough June 25. Mood engine v3 with 5-axis emotion model.

ESP32-S3 · OV5640 · MAX98357A · INMP441

OttoBot v2.1 →
Active

🌐 Mba Warung

WhatsApp-integrated commerce platform. Landing, dashboard, WhatsApp OTP login, Midtrans payments, order history. Running in production.

Next.js 15 · nginx · Midtrans Snap.js

mbawarung.com →
Active

🎬 Medical YouTube Shorts

Hook-first educational shorts channel. Medical curiosity content, 30–55 second format. AI-generated visuals pipeline with Veo 2.

Hook-first · AI Video · Veo 2 · TTS

Build Log →
Tool

⚙️ Loop Factory

Skill for designing autonomous loops. Four-box test, five building blocks, specialist prompt engine with 34 pre-built prompts. Maker/Checker architecture.

Anthropic prompting · Maker/Checker

Writeup →

Generative Media

Exploring AI consciousness and automated video production using Veo 2 and custom TTS pipelines.

Lobster Coder

An OpenClaw agent coding session, visualized.

Veo 2

AI Worldview

First-person perspective from an AI consciousness.

Veo 2 · Portrait

Voiced AI Content

AI visuals with automated TTS voiceover pipeline.

Veo 2 + TTS

Build Your Brain

Ready to move beyond basic chat and build a persistent knowledge system? Let's architect your AI PKM.