CognitiveXCOGNITIVEX · THE LARGE COGNITION MODEL
Your AI forgets. Oursremembers, learns, and evolves.
CognitiveX is the Large Cognition Model, a cognitive layer that gives any app or agent living memory. It remembers what happened, learns your patterns, and gets smarter with every interaction. The memory is the model.
Free to start, no card required. Building an app instead? Read the docs →
Query → Living Memory → Reasoning → Learning → Evolution, and back4-tier memory: semantic · episodic · procedural · foundationalpattern detection · salience · overnight dream consolidationcross-agent recall over MCP
THE PARADIGM SHIFT
Not another LLM wrapper.
An LLM does Query → Model → Response → forget. The LCM closes the loop: Query → Living Memory → Reasoning → Learning → Evolution. Storage and search are table stakes. Everything below the line is where the category begins.
| Capability | LLMs | RAG | Vector DBs | CognitiveX LCM |
|---|---|---|---|---|
| Stores information | ||||
| Semantic search | ||||
| Learns from usage | ||||
| Detects patterns | ||||
| Makes predictions | ||||
| Self-improves | ||||
| Remembers context | partial |
THE COGNITION LOOP
The memory modifies itself. That's cognition.
A model answers and forgets. The LCM runs a recursive loop: every interaction is remembered, reflected on, reasoned over, decided, and learned from, and that learning rewrites the memory the next query reads. The substrate that answers you today is shaped by everything it answered before.
- 1Memoryevery interaction is captured into the living substrate
- 2Reflectionit revisits what it holds and consolidates
- 3Cognitionit reasons over the patterns, not just the hits
- 4Decisionit surfaces the answer and the next move
- 5Learningthe outcome rewrites the memory the next query reads
THE COGNITION ENGINE
A living substrate, not a database.
Underneath the API is a cognitive substrate that decays, consolidates, and reorganizes itself, an orchestra of algorithms playing in concert. We keep the score private; what you see is the music.
Living memory
- not a store, but a substrate that decays, consolidates, and strengthens with use
- 4 tiers: semantic, episodic, procedural, foundational
- overnight dream consolidation turns scattered events into patterns
An algorithm orchestra
- many specialized algorithms working in concert, not one model call
- pattern detection and salience decide what surfaces, and when
- reflection and introspection let it reason about its own state
Cognitive modes
One dial for depth of reasoning vs. cost, set per query or left to adapt.
BUILD ON THE LCM
An SDK that speaks human.
Give any app or agent living memory in a few lines. The cogx SDK, MCP server, and HTTP API all front the same cognition engine: remember, recall, detect patterns, reflect. One install wires it into Claude, Cursor, Codex, or your own backend.
- SDK ·
@cognitivx/sdk - MCP ·
api.cognitivx.io/mcp - HTTP API +
cogxCLI
import { configureApiClient, memory } from "@cognitivx/sdk"; // remember, with the reason, not just the factawait memory.remember({ content: "chose Stripe over Paddle for lower MoR friction" }); // recall returns the why and the next moveconst hits = await memory.recall({ query: "our payments decision", deep: true }); // talk lets the cognition engine surface patternsconst reply = await memory.talk({ message: "what patterns do you see?" });const state = await memory.reflect();
@cognitivx/sdk surface.ICOG · THE FLAGSHIP
iCog, our personal AI, runs on the LCM.
iCog is the consumer app we built on CognitiveX, proof the cognition engine works at the surface. It greets you by what it already knows, carries your context across every AI tool over MCP, and gets sharper the more you use it. Same memory, every assistant.
PROOF
Memory that actually holds.
We evaluate recall on LongMemEval, the long-horizon memory benchmark, instead of grading ourselves on a demo. The honest version, how the LCM stacks up against memory layers that only store and retrieve, lives on the comparison page.
YOUR DATA, YOUR CONTROL
Private by default.
- you control what's kept
- export or delete anything, anytime
- private by default, never sold, shared, or used to train shared models
PRICING
Start free. Scale when your app gets smart.
- 100 recall credits / month
- 1,000 memories
- SDK, MCP, and HTTP API
- 500,000 recall credits / month
- 1,000,000 memories
- dream consolidation + deep recall
- 15,000,000 recall credits / month
- unlimited memories
- highest-priority recall
- unlimited memories
- credits billed as you use them
- set your own spend cap
A recall costs credits by depth: Foundational 1 · Standard 3 · Deep 10. Storing memories is free. How CognitiveX compares →
QUESTIONS
What is a Large Cognition Model?
An LCM is a cognitive layer that sits around the LLM. Where an LLM does query → model → response → forget, the LCM adds living memory, reasoning, and learning that compounds: it remembers what happened, detects patterns, and gets sharper with every interaction. The memory is the model.
How is it different from an LLM?
An LLM is stateless. Each call starts from zero plus whatever you paste into the prompt. The LCM keeps a persistent, evolving memory and runs a recursive loop over it, so behavior carries across sessions, agents, and tools. We treat the LLM as infrastructure; the intelligence lives in the cognition engine.
How is it different from mem0 or Zep?
Most memory layers store and retrieve. They remember text and search it. CognitiveX consolidates and learns: episodic events get promoted to semantic facts, patterns surface on their own, and overnight dream consolidation reorganizes the substrate. The honest, side-by-side version is on the comparison page.
Which tools does it work with?
Any MCP client (Claude Desktop, Claude Code, Cursor, Codex, ChatGPT), plus the cogx SDK and HTTP API for your own apps and agents. One install (cogx mcp install) wires it in.
Where is my data, and who can see it?
In your account, in encrypted databases. It's private by default, never sold, shared, or used to train models other people can access. You can export or delete any of it at any time.
What is iCog?
iCog is our flagship consumer app, a personal AI built on the CognitiveX LCM. It's the easiest way to feel what living memory does; developers build their own products on the same engine via the SDK.