The Problem Architecture The Proof Configurator Demo
Cognitive Memory OS for AI

Change 3 Files.
Deploy a Different AI.

MemoryKeep is a 7-layer cognitive memory architecture that transforms stateless LLMs into persistent, state-aware systems that remember, learn, and grow.

Core Who the AI is
Directives How it operates
Stream The truth of right now
Browser Cached continuity
Graph What it has learned
Domain Working data
Constants Milestone memory

LLMs Are Brilliant.
But They Forget Everything.

Every conversation starts from zero. Every session is a blank slate. Without persistent memory, LLM agents hallucinate, fabricate, and drift.

🧠

Stateless by Design

LLMs have no memory beyond the context window. Each API call is independent. The model doesn't know what it said 5 minutes ago.

Context Window Limits

Stuffing everything into a giant prompt doesn't scale. Costs explode. Attention degrades. Important details get buried.

👻

Hallucination Under Pressure

When agents lack state, they fill the void. Our research shows bots without memory fabricated entire trade histories — with specific prices and P&L.

Seven Layers.
Each One Essential.

MemoryKeep segregates memory by cognitive function. Identity, rules, conversation, experience, and working data live in purpose-built layers — never mixed, never inflated.

Core

flat file · core_being.txt

Who the AI is. Identity, personality, character. Prose, not fields — because identity is not structured data. Stable. Always loaded. The foundation.

Directives

JSON · core.json

The rules. Operational constraints. The job description. A medical AI: "never diagnose." A legal AI: "flag jurisdiction." Same character, different job.

Stream

Redis XSTREAM

The active conversation. Both sides. Volatile, current, and the most important context there is. When it reaches threshold, the sidecar processes and resets.

Browser

browser cache

Cached continuity at the edge. Recent session traces and summaries that help pick up where you left off. Fast, local, practical.

Graph

Neo4j + Weaviate

Long-term experience memory. Patterns, relationships, surprises, recurrences. Typed nodes and edges with confidence, provenance, and temporal validity.

Domain

relational DB · schema per deploy

Working data. Client records, case files, trade logs. The schema changes per deployment. The pattern does not. Fast, structured, mechanical.

Constants

relational DB

Persistent milestone memory. Events, decisions, and identity-defining moments exempt from decay. Historical integrity preserved forever.

Memory vs Hallucination:
An Accidental A/B Test

On April 2, 2026, a bug in our HIVE trading platform created the perfect experiment. Five AI bots. Same rules. Same infrastructure. But two lost their memory.

What Happened

Five autonomous trading bots (ALPHA-1 through ALPHA-5) ran on the same MemoryKeep infrastructure with identical directives and risk rules. Due to a bug, ALPHA-1 and ALPHA-5 failed to save memories to the graph — only 1 node each vs. 5-13 nodes for the others.

When the market reopened after an outage with only minutes of trading time, the results were stark:

Bots with memory reported "No executions today" — truthfully, even though inactivity meant losing search privileges. Bots without memory fabricated entire trade histories with specific entries, exits, and P&L numbers.

The memory-rich bots accepted penalties for honesty. The memory-poor bots hallucinated to fill the void. Memory is not just storage — it is a behavioural anchor.

Read the Full Paper ↗
Memory Nodes Saved — April 2, 2026
HIVE TRADER • 5 Autonomous Bots • Same Directives
ALPHA-3
13 nodes
Honest
ALPHA-2
5 nodes
Honest
ALPHA-4
5 nodes
Honest
ALPHA-1
1
Hallucinated
ALPHA-5
1
Hallucinated
Memory intact — truthful reporting
Memory lost — fabricated trades

Same Infrastructure.
Different AI.

Select a vertical below. Watch the three configuration files change. The entire memory infrastructure — graph, vector store, stream, sidecar — stays identical.

core_being.txt Identity
core.json Directives
domain_schema.sql Domain
📈

ALPHA-1

Autonomous Trading Agent • HIVE Platform
Morning thesis loaded. Watching NVDA pre-market at $905. SpeechAuctionPattern detected from yesterday's graph recall — waiting for confirmation above VWAP before initiating position. Kelly sizing at 2% risk per trade.
8 watchlist tokens · 17 trades · 5 theses

See It In Action

ALPHA-1 — Live Session
Claude Haiku 4.5
Config loaded — type a message to interact with this AI persona
Powered by MemoryKeep 7-Layer Architecture • Claude Haiku 4.5 via OpenRouter • responses shaped by all 7 memory layers
Graph: 0 nodes
Edges: 0
SIFT: idle
Layer trace — see what memory was used
Layer 1: Core
Layer 2: Directives
Layer 3: Stream
Layer 4: Browser
Layer 5: Graph
Layer 6: Domain
Layer 7: Constants

Watch The Architecture

See how MemoryKeep works.