mama-dev

MAMA OS — Local AI Runtime with Connected Memory

Your AI remembers why, not just what.

MAMA OS is a local AI runtime that turns conversations, decisions, and work artifacts into a structured knowledge graph. Every choice is saved with reasoning. Every mistake becomes a reference. When AI agents connect to MAMA, they don't just answer — they understand your context and recommend with evidence.

Session 1:  "Use SQLite for the project"
Session 5:  "Switch to PostgreSQL — SQLite can't handle concurrent writes"
Session 12: "Should we use MongoDB for the new service?"
            → MAMA: "You switched from SQLite to PostgreSQL in Session 5
               because of concurrent write issues. MongoDB has similar
               trade-offs. Here's what happened last time..."

Why MAMA OS

AI agents today are brilliant but amnesiac. They forget everything between sessions. MAMA solves this:

	Without MAMA	With MAMA
Memory	Forgets after session	Knowledge graph with decision evolution
Context	Generic answers	Answers grounded in your history
Mistakes	Repeat the same errors	Past failures prevent future ones
Knowledge	Only general knowledge	Domain-specific, accumulated over time

Why We Pivoted

MAMA started as a multi-agent dashboard — skills marketplace, playground, agent swarm, cron scheduler. Those features still exist and work, but they're no longer the focus.

Why? Because AI agents commoditized all of it. Claude, GPT, Codex can all orchestrate sub-agents, execute code, and run tools out of the box. Building another agent framework is competing with the foundation model providers themselves.

What AI agents cannot do is remember. Every session starts from zero. That's the gap only a persistent memory layer can fill — and that's where MAMA's real value lies.

Commoditized (still works, not the focus)	Unique to MAMA (core focus)
Multi-agent orchestration	Knowledge graph with decision evolution
Skills/plugins marketplace	Automatic memory extraction from conversations
Code sandbox	Cross-source intelligence (code + chat + email)
Cron scheduling	Domain-specific recall with "why" context

The Vision: From Tool to Partner

Today:     General-purpose AI agent (smart but no context)
+ MAMA:    AI agent with your domain knowledge (smart + experienced)
Future:    AI that understands why your team makes decisions the way it does

Tacit knowledge — the know-how that lives in conversations, failed experiments, and tribal memory — is the most valuable and hardest to capture asset in any organization. MAMA captures it automatically through a knowledge graph that preserves not just facts, but the reasoning behind every decision.

Architecture