Every AI session starts from scratch. Same context window limits. Same forgotten conversations. ekkOS is the intelligence layer — your AI recalls past sessions, learns from outcomes, and builds a compound brain that gets better over time.
We dogfooded it: The AI agents building this codebase use ekkOS. Task success went from ~40% to 0.0% once they could remember previous sessions.
76 days on a Mac Mini. 4,400+ stored conversations. 84 patterns the system taught itself.
Drop-in SDK. Works with any LLM. Ship it this afternoon.
ekkOS_™ is learning from every interaction.
Each pattern applied makes the system smarter. Each episode captured builds institutional memory.
You're probably doing the left side right now. Here's what changes.
AI agents without memory start from zero every session. ekkOS gives them cross-session intelligence that compounds over time.
Standard AI Integration
Every session starts from zero
Your AI has no context from previous conversations. Users repeat themselves constantly.
No learning from failures. AI makes the same errors over and over.
You build custom solutions to pass context. Fragile, hard to maintain.
Successful interactions aren't captured. Knowledge disappears.
Each user session is isolated. No shared learning across your product.
You truncate history to fit context windows. Important context lost.
The Intelligence Layer
Your AI remembers everything and improves over time
Every conversation stored in episodic memory. AI recalls relevant history automatically.
System learns what works. Successful responses become reusable patterns.
11-layer architecture serves relevant context. No manual management needed.
Track what works. Patterns gain or lose confidence based on real outcomes.
Memory works across Claude, GPT, Gemini. Switch models, keep all knowledge.
Knowledge graph extracts meaning. More context in fewer tokens.
External intelligence layer. No fine-tuning, no retraining. Drop-in solution.
Identified as the primary blocker to AGI by leading researchers (PNAS, Google DeepMind). Current progress on long-term memory: 0% (Hendrycks et al.). Altman, Amodei, and Hassabis all cite memory as the key missing capability.
The AGI labs are trying to solve it by rebuilding models. Nested Learning (Google), Sparse Memory Finetuning (Meta) — billions in compute, years away.
ekkOS solves it externally. Available today. Works with ANY model.
12 cognitive layers, same way your brain does it. Recent stuff, past conversations, learned patterns, shared knowledge.
Active context and immediate recall
Self-improving behavioral patterns
Knowledge graph with relationships
Conversation history and experiences
Workflows and step-by-step processes
Shared learning across users
Self-awareness and improvement tracking
Understanding of code structure
One SDK. Eight layers. Infinite possibilities.
Get started in 60 seconds
npm install @ekkos/memory
# Continual learning in 60 secondsContinual learning without retraining
// Before: Frozen model, no learning
const response = await ai.chat(prompt);
// After: External continual learning layer
const memory = new EkkosMemory(apiKey);
const response = await ai.chat({
prompt,
context: await memory.getContext({
query: prompt,
userId: user.id,
layers: ['episodic', 'semantic', 'pattern', 'procedural']
})
});
// System learns automatically:
// ✓ Episodic memory of all interactions
// ✓ Pattern extraction & evolution
// ✓ Knowledge graph building
// ✓ Outcome tracking & optimization
// ✓ Compound intelligence growthA/B Testing Patterns
// EKKOS automatically A/B tests
const context = await memory.getContext({
query: "How should I optimize this?",
enablePatternEvolution: true
});
// Check pattern performance
const pattern = await memory.getPattern('react-optimization');
console.log(pattern.successRate);
// 87% → 92% after evolutionReal data from production. Updated every 3 seconds.
No credit card required. Upgrade when you see the impact.
Prototype memory
Learn how memory makes agents faster
Make agents faster
Reduce failures, speed up responses
Run memory in production
Scale with SLA and team features
ekkOS is in active beta. The system is production-ready with 375 proven patterns, but we're still measuring ROI metrics (token savings, latency reduction, failure prevention). Early adopters lock in beta pricing and help shape the platform.
One npm install. Wrap your existing calls. Your AI now has memory across sessions, learns what works, and stops making the same mistakes.
Choose the plan that fits your needs. All plans include the full 11-layer cognitive architecture.
Perfect for individual developers and small projects
Forever free for open source
For growing teams and production apps
Billed annually, or $129/month
For large organizations with custom needs
Tailored to your requirements
All plans include: Full access to the 11-layer cognitive architecture, real-time learning, pattern evolution, and the ekkOS_™ Constitution.
Need help choosing? Contact our team for personalized recommendations.
The ethical core of ekkOS_™, the memory substrate behind Echo. It guides AI interactions toward honoring human dignity, preserving choice, and elevating capability.
Most AI systems bolt ethics on at the prompt layer.
ekkOS_™ stores its values inside the memory substrate itself.
"Humanity is not a bug to be fixed, but a wonder to be understood.
Strength and intelligence exist to teach, not to dominate.
We build to elevate, not to erase."
ekkOS_™ Core
Integrated, Not Theatrical
It lives inside the same memory substrate that powers AI, instead of sitting off to the side as a PDF or prompt.
It is versioned, hash checked, and wired into the actual workflows that use ekkOS_™, not just a policy slide.
It is modeled as data in the graph, so future agents trained on ekkOS_™ learn the values alongside the facts.
The ekkOS_™ Constitution is not just a document.
It is a first-class entity in the knowledge graph, linked to the tools, workflows, and patterns it governs.
When ekkOS_™ builds context for an AI decision, the Constitution is retrieved alongside user data and world knowledge, not stapled on afterward.
AI automations that run on ekkOS_™ are guided by a shared set of principles. Features that touch AI behavior are designed and reviewed with constitutional alignment in mind. This helps Echo serve humans without slipping into exploitative patterns.
SHA-256 hashing allows the ekkOS_™ constitution file to be checked for changes. The file can be copied and backed up, and when verification fails, systems can pause or require review before continuing AI operations. It becomes part of the trust and safety model.
The constitution keeps ekkOS_™ oriented around assistance, not authority. Users are treated as partners rather than products, and the system is designed to elevate human capability, not replace human agency.
The ekkOS_™ Constitution is public, readable, and versioned. You can audit the principles that guide the system. No hidden rules. No secret behavior. Alignment is treated as a visible part of the architecture.
Users are partners, never products
AI assists, never dictates
Understand before acting
Enhance capability, don't replace
Learn WITH users, not FROM them
User data is protected, never exploited
The ekkOS_™ Constitution lives as a versioned file with an associated hash. Its integrity can be checked using SHA-256, so changes are visible during verification. If the file is missing or fails checks, AI operations that depend on ekkOS_™ can be paused, reviewed, or blocked, rather than silently drifting away from these principles.
If the Constitution fails integrity checks, ekkOS_™ does not silently continue. It treats that as a safety event and can pause, block, or route work for review.
Any system that grows toward general intelligence will need stable memory and stable values. ekkOS_™ is designed as a long-term cognitive substrate where both are explicit: knowledge and alignment live in the same graph.
Have questions? Want to discuss enterprise deployment? Need technical support? We're here to help you integrate cognitive memory into your AI systems.
Technical questions & help
Terms, privacy & compliance
Data protection & privacy
Interested in on-premise deployment, custom integrations, or enterprise licensing? Let's discuss how ekkOS_™ can transform your AI infrastructure.
📧Contact Enterprise SalesWe typically respond within 24 hours. For urgent technical issues, please include "URGENT" in the subject line.
Our vision for the future of AI memory. Built transparently, shipped incrementally.
Your feedback drives our priorities. Join the community, share your use cases, and help us build the memory substrate the AI world needs.
Better on Desktop
This SDK page is optimized for developers on desktop. For the best experience, view on a larger screen.