Product Pricing Why Now About Get Early Access →
Character Intelligence Middleware

We build the
minds
behind
your NPCs

Cognition Labs creates the character intelligence infrastructure for the games industry. Our first product gives non-player characters the ability to remember, think and grow.

Flagship Product
Cortex NPC
●  Coming Soon

Persistent memory, goal-driven behaviour and dynamic dialogue for your game characters — without rebuilding your engine. Available soon as Unity and Unreal plugins.

7-Layer
Runtime Architecture
Unity + UE
Day-One Engine Plugins
∞ Memory
Persistent NPC History
Persistent NPC Memory
Goal-Driven Behaviour
Dynamic Dialogue
Unity & Unreal Plugins
Engine-Agnostic Runtime
Social Relationship Graphs
Deterministic Policy Engine
Studio-Defined Constraints
Persistent NPC Memory
Goal-Driven Behaviour
Dynamic Dialogue
Unity & Unreal Plugins
Engine-Agnostic Runtime
Social Relationship Graphs
Deterministic Policy Engine
Studio-Defined Constraints
01

Scripted NPCs
don't scale

The tools studios have used for decades — state machines, branching dialogue trees and one-off trigger logic — create structural problems that compound as games grow larger and live longer. Every new line, quest branch or player reaction requires manual authoring and QA. That is not a content pipeline. That is renting time from your writers indefinitely.

Issue 01
Characters feel repetitive

NPCs exhaust their script within hours, shattering immersion and discouraging replay.

Issue 02
Every line needs authoring

New quest branches, reactions and dialogue all require writer time, build cycles and QA.

Issue 03
Live service demands more

Open worlds and GaaS titles need reactive systems that static pipelines simply cannot provide.

Issue 04
No personalisation at scale

Studios cannot tailor NPC behaviour to individual player history when everything lives in hard-coded trees.

02 — Our Flagship Product
CortexNPC

Not a chatbot inside a game. A multi-layer character intelligence runtime with deterministic guardrails.

Each NPC carries its own memory of what it saw, heard and experienced. It knows its goals, its loyalties, its emotional state and its relationships. When the player interacts, the system retrieves relevant memories, passes structured context to a behaviour planner, and returns a response that includes animation intent, emotional tone, objective and priority — not just raw text.

The outcome is written back to memory. So the tavern keeper who was robbed three hours ago, knows the town guard is looking for suspects, and also remembers the player once saved his sister — greets the player with genuine suspicion, but decides not to report them. Immediately.

Studio-defined templates, tags and policy limits ensure the platform generates strictly within your lore and content boundaries. You stay in control. The characters come alive.

How It
Works

01
Structured context in
The game sends nearby actors, quest state, time of day, player reputation, local inventory and design constraints via plugin or API.
02
Rules layer filters
A lightweight rules layer filters what the NPC is allowed to know or do, preventing lore-breaking, impossible actions and unsafe outputs.
03
Memory retrieval
A memory service retrieves the NPC's persistent memories, recent interactions and relationship data using hybrid structured and semantic search.
04
Behaviour planner decides
A planner decides the most relevant response type: speak, refuse, flee, offer help, continue current task, notify another NPC, and so on.
05
Structured output returned
The response returns to the game as structured output: chosen animation intent, emotional tone, objective, target and priority — not just raw text.
06
Memory written back
The system writes the outcome back to memory. The NPC changes over time. It grows. It remembers. It surprises.
M
Persistent Memory

Every NPC tracks what it saw, heard and experienced across sessions. Memory is hybrid structured and semantic — so retrieval is both fast and contextually aware.

P
Personality & Emotion

Maintains goals, loyalties, emotional state and social relationships that evolve naturally through every player interaction.

T
Tiered Intelligence

Tune cost, latency and autonomy per-NPC. Not every character needs full intelligence at all times. Use lightweight local models for ambient characters.

D
Developer Tooling

Full tracing, simulation harness, trace viewer and personality tuning screens. Debug exactly why a character said something or changed behaviour.

E
Engine Agnostic

Unity and Unreal plugins are the fast path. The platform also exposes a C++ SDK and REST, WebSocket and gRPC interfaces for custom and proprietary engines.

C
Studio Control

Generate strictly within studio-defined templates, tags and policy limits. Lore boundaries, safety rules and approval workflows always remain in your hands.

Pricing

Start free, scale as you ship. Enterprise deals available for AAA publishers, platform holders and studios on custom engines.

Indie
Free
up to usage cap
  • Cloud SaaS plugin
  • Unity & Unreal packages
  • Up to 3 NPC characters
  • Community support
  • Developer dashboard
Request Access
Indie Pro
£79–499
per month
  • Everything in Free
  • Unlimited NPC characters
  • Higher usage limits
  • Analytics & trace viewer
  • Email support
Request Access
AA Studio
£15k–100k
per year
  • Production SaaS & support
  • Managed deployment option
  • SLAs & predictable terms
  • Solutions engineering
  • Simulation sandbox
Request Access
AAA / Enterprise
Custom
private cloud or on-prem
  • Private cloud or on-prem
  • Source-level integration
  • Custom engine support
  • Dedicated account team
  • Security & compliance
Request Access

Why
Now

Several structural forces are converging to make character intelligence middleware not just possible, but essential for competitive game studios.

01
Generative models are ready
Foundational models can now generate convincing, in-character language reliably enough for production games.
02
Studios need more with less
Content demands are outpacing headcount budgets. AI-assisted character systems directly reduce the authoring burden.
03
Open worlds are live services
Modern open-world and GaaS titles benefit disproportionately from reactive, persistent character systems.
04
Players expect emergence
Player expectations are shifting from static content trees toward emergent, systems-driven world experiences.
05
Middleware markets are huge
Gaming middleware becomes enormously valuable once embedded in production stacks — see Havok, FMOD, Wwise.
06
The category is being defined now
The tooling that becomes standard is built in the next 24 months. The window to own the category definition is open.

About
Cognition Labs

Cognition Labs is a UK-based technology company building character intelligence infrastructure for the games industry. Our long-term ambition is to become for character intelligence what Havok became for physics — a standard middleware layer adopted across the industry.

We believe the next generation of great games will not be defined by better graphics or larger maps. They will be defined by characters that genuinely remember you, react to you and surprise you.

This is not a bet on one hit game. It is a bet on the infrastructure that makes future games feel dramatically more alive.

Our Principles

01
Production first
We build for shipped games, not demos. Latency, reliability and cost-per-interaction matter as much as raw model quality.
02
Studio control is non-negotiable
Studios define the rules. We generate within their boundaries. Lore, safety and design constraints always win.
03
Start small, prove fast
The best first proof is not a thousand autonomous villagers. It is a small set of important NPCs that noticeably improve the experience.
04
Build the moat through tooling
The strongest defensibility is embedded integrations, production tooling and battle-tested performance in shipped titles.

Bring your
characters
to life

Join the early access programme and be among the first studios to integrate Cortex NPC into your game.