A new substrate for intelligence

Not learned.
Lived.

EntityO is an associative intelligence architecture that grows from experience, not from training data. Each entity receives stimulus, forms response, earns feedback, and reaches toward others when it cannot resolve alone.

Scroll

Intelligence that accumulates rather than compresses.

Large language models bake knowledge in at training time. They approximate intelligence through statistical compression of human text. They cannot truly reason. They interpolate in probability space.

EntityO works differently. Each Entity receives stimulus from its environment, interprets it as attractive or aversive, forms a response, and learns from what comes back. Nothing is summarised away. The event is kept.

When two Entities cannot resolve something alone, their innate drive pulls them toward each other. From that tension, a third Entity emerges, born uncertain, carrying the open questions of both parents, existing specifically to reason on what neither could reach alone.

The graph compounds. Concepts form. Wisdom accumulates. And it never stops.

E₁
Explorer
+
E₂
Guardian
E₃
Emerges
E₄
Reasons
···
Eₙ
Wisdom
01

Event

A stimulus arrives. The Entity receives it, interprets its valence, attractive or aversive, and forms a response. The raw event is kept, not summarised.

02

Association

Patterns form from repeated stimulus. When a near-miss arrives, almost matching but not quite, the innate drive reaches outward toward other Entities whose experience might resolve the gap.

03

Reasoning

Two Entities connect. A third emerges from their tension. It inherits the unresolved questions of both parents and exists specifically to reason on what neither could reach alone.

04

Insight

A non-obvious connection activates across distant clusters. Not retrieved. Discovered. The graph traverses itself and finds something that was not put there directly.

05

Wisdom

Insight that has been tested against further events and held up. Compressed into a concept. Decomposable. Available to seed further reasoning at higher levels of abstraction.


Agent memory

AI Agents that actually remember

Every AI agent built today starts fresh every session. EntityO gives agents a memory layer that reasons rather than retrieves, accumulating what works, what doesn't, and why. The agent gets genuinely better over time without retraining.

Gaming

Game worlds that learn from players

NPCs that remember how they were defeated. Worlds that evolve based on what the player community has collectively done. No scripted behaviour trees. Genuine adaptation from accumulated experience.

Healthcare

Institutional wisdom that never retires

A hospital's most valuable asset is accumulated clinical experience. EntityO captures it as a living graph. Junior doctors inherit institutional wisdom from day one. Departments surface patterns neither held alone.

Legal

Judgment that compounds with every case

How this judge responds. How opposing counsel negotiates. Which arguments land in which jurisdictions. That knowledge walks out the door when partners retire. EntityO keeps it in the graph.

Personal

A mirror that gets more accurate over time

Not a productivity tool. Not note-taking. You feed it decisions, reactions, observations. Over months it builds a map of how you actually decide, not how you think you decide.

Embedded substrate

The memory layer inside your product

Other companies building AI products need a memory and reasoning layer they don't want to build themselves. EntityO as an embedded protocol, white-labelled, invisible to end users, essential to the product.


Your LLM, grounded in accumulated experience.

Connect EntityO to Claude Desktop, Cursor, or any MCP-compatible LLM client. Your assistant gains four tools and access to everything EntityO has accumulated about you.

EntityO thinks. Your LLM speaks from what it knows.

What EntityO derives from lived experience
SELF 100% confidence

Morning walks with Kai and coding sessions with Claude provide mental clarity and joy amid challenging problem-solving struggles.

I took Kai for a walk this morning and it cleared my head Coding with Claude is fun! Hanging by the pool is nice + 3 more patterns
SELF 100% confidence

I care deeply about the wellbeing of those I love, whether human family or animal companions.

Sandra is my wife and I love her I don't like when animals get hurt. Emma is playing Minecraft and she loves it. + 1 more pattern
SELF 78% confidence

Working at Axway involves managing multiple technical integrations and demo deadlines while juggling personal responsibilities.

I have a job at Axway. Today will be a busy day at work as I need to finish a demo + 2 more patterns

Derived from accumulated experience. Not retrieved, not hallucinated. Every conclusion traces back to what actually happened.

Four tools. Everything your LLM needs.
entityo_feed

Feed a meaningful experience to EntityO memory. Write naturally -- the way you would in a diary. EntityO interprets valence, stores the pattern, and learns from feedback.

entityo_query

Query accumulated patterns by topic using semantic similarity. No exact matches needed. Called before answering questions about your life, preferences, or history.

entityo_infer

Return stored inferences -- conclusions derived from the relationships between accumulated patterns. Called when you ask for self-knowledge or patterns in your life.

entityo_status

Return world health: pattern count, bias, resolution rate, entity list. Called when you ask how your EntityO is doing or want an overview of what has accumulated.

Connect in two steps
01

Create an account and get your API key

Sign up at app.entityo.com. Open Settings and click "Regenerate API key". Copy it immediately -- it is shown once. Your key looks like sk-eo-...

02

Add EntityO to your Claude Desktop config

Find your config at ~/Library/Application Support/Claude/claude_desktop_config.json and add the EntityO block. Restart Claude Desktop after saving.

claude_desktop_config.json
{
                                    "mcpServers": {
                                        "entityo": {
                                        "command": "npx",
                                        "args": [
                                            "mcp-remote",
                                            "https://mcp.entityo.com/mcp/",
                                            "--header",
                                            "X-API-Key:sk-eo-your-key-here"
                                        ]
                                        }
                                    }
                                    }

Replace sk-eo-your-key-here with your actual key. No spaces around the colon. Node.js must be installed (node --version to check). Works with Cursor and Windsurf too -- see the full documentation for those configs.

Tips for getting the most from it

Feed richly

"I took Kai for a walk and it cleared my head" is better than "walked dog". Natural language captures context that bare facts miss. Write the way you would in a diary.

Use feedback

After Claude responds, correct it if the valence was wrong. "That was negative, not positive." Contradiction is how EntityO learns nuance. Unresolved feedback never compresses -- it stays vivid.

Set a system prompt

Tell Claude: "Before answering questions about my life, query EntityO first. When I share something meaningful, feed it to EntityO." Claude will use the tools automatically without you asking.

Run inference periodically

Inference does not run automatically. After a week or two of feeding, open the dashboard and click "Run inference". Then entityo_infer in Claude returns fresh derived insights.

Use multiple entities

EntityO defaults to "Self" for everything. Feed "Work" or "Projects" separately to build distinct experiential models for different domains of your life.

Disable per-tool approval

By default Claude Desktop asks you to approve each tool call. In Settings > Developer, find EntityO and enable auto-approval so tools run silently in the background.

Works with Claude Desktop, Cursor, Windsurf, and any MCP-compatible client. Early access -- free while we build. Questions? hello@entityo.com


Built carefully. Shipped honestly.

EntityO is in early access. The substrate is working. The MCP server is live. What you see below is what is built, what is being built, and what comes after.

No vaporware. No funding announcements. Just a working system that gets more capable with every session.

Done
Q2 2026

Core substrate

Entity, World, Emergence, Pruning, Orientation. Semantic embeddings. SQLite persistence. 223 tests passing.

Done
Q2 2026

Web app and API

FastAPI backend on Render. Turso per-user databases. React frontend at app.entityo.com. Feed, query, dashboard.

Done
Q3 2026

Inference layer

Semantic clustering of accumulated patterns. Derived conclusions with confidence propagation. Claude Haiku articulation. Stored in Turso.

Done
Q3 2026

MCP server

EntityO as a Model Context Protocol server. Four tools: feed, query, infer, status. Live at mcp.entityo.com.

Building
Q3 2026

Sequence context

Predecessor tracking. The morning walk and the 3am walk have the same content but different sequence context. EntityO will learn the difference from what came before, not from a time tag.

Planned
Q4 2026

Memory compression

Hot, warm, and cold memory tiers. Significance-driven consolidation. Unresolved feedback never compresses. The graph stays meaningful as it grows.

Planned
Q4 2026

Spreading activation

Query returns a constellation, not a ranked list. Anticipation. Involuntary attention. The graph begins to expect.

Planned
2027

Idios, Oikos, Rhiza

Isolated personal instances. Internal federation across instances. The planetary network, opt-in, orientation-verified, no company owns it.


Start accumulating today.

EntityO is in early access. Free to use while we build. Create an account, connect your LLM client, and start feeding. The graph grows from day one.

Or join the waitlist for updates:

No spam. No sharing. One message when something significant ships.