Architecture
ARQERA is not an app. It is the first AI-native operating system — the intelligence layer that sees, decides, and acts across your entire operation.
The AI-Native Operating System
ARQERA is an operating system where the AI IS the OS. Screens are ambient surfaces that show what the intelligence is doing. Every capability is built internally first, then made available to customers.
AI IS the OS
Ore is the primary interface. Voice, text, ambient intelligence. Features are Spaces within the OS, accessed by asking Ore, navigating the Mycelium, or pressing a command palette.
Mathematical Biology
The architecture draws on biological systems — Physarum slime mould dynamics, Bayesian trust networks, and entropy theory. These are not metaphors. They are the routing and decision algorithms.
Self-Consuming
ARQERA uses what it sells. Every capability — compliance, analytics, governance, cost optimisation — is deployed internally first. If it does not work for ARQERA, it does not ship to customers.
The Four AIs
Named in Yoruba, four specialised intelligences work together. Each has a distinct role in the operating system.
Oju
Eye — Sees
Continuous monitoring and anomaly detection across your entire operation. Not a dashboard — a living awareness that surfaces what matters before you ask.
Ori
Head — Thinks
Strategy, risk assessment, and planning. Connects dots across every tool you use, sees patterns before they become problems, and recommends the optimal path forward.
Ore
Friend — Helps
Your AI companion and the primary interface to the OS. Ore is the surface you interact with — voice, text, ambient intelligence. It wakes you with what is done, not what is due.
Ara
Body — Acts
The execution engine. Ara carries out actions across all your integrations — sending emails, creating tickets, deploying code, running workflows. Governed, audited, always accountable.
How They Work Together
Oju
Detects
Ori
Evaluates
Ore
Presents
Ara
Executes
An anomaly is detected by Oju. Ori evaluates the risk and recommends a course of action. Ore presents the recommendation to the user (or auto-proceeds if trust is high enough). Ara executes the approved action across the relevant integrations. Every step emits a signed evidence artifact.
The Gravity Architecture
Six mathematical principles drawn from biology, economics, and information theory govern every decision the system makes.
Physarum Dynamics
Inspired by biological slime mould networks. Routes that succeed are strengthened. Routes that fail naturally decay. The system continuously optimises itself — no manual tuning required.
Bellman Optimality
Every decision has a computed cost — factoring risk, time, and effort. Policy violations carry infinite cost, making them impassable barriers. The system always finds the lowest-cost valid path.
Bayesian Trust
Trust is not binary — it is probabilistic and evidence-based. Trust accrues through successful outcomes and decays without evidence. Recent results are weighted more heavily than historical patterns.
Network Value
Knowledge compounds super-linearly. Every fact added to the system makes all existing facts more valuable. The platform gets smarter with every tenant, every action, every outcome.
Hexagonal Coverage
Every feature requires six neighbours to be complete: tests, security, user experience, documentation, performance, and observability. Gaps are detected and flagged automatically.
Entropy Self-Organisation
When something breaks, the system does not retry the same path. It self-organises toward order by routing around failures and adapting through local rules — like a living organism healing a wound.
OS Layers
The operating system is structured in three layers: The Mycelium home surface, 12 App Spaces, and Ore as the ambient AI chrome.
The Mycelium
The home surface. A dark bioluminescent field with flowing tendrils that connect ambient nodes. Each node represents a Space. The tendrils visualise conductance paths — the biological routing of the Gravity Architecture, made visible.
Ore (OS Chrome)
A floating morphing pill anchored at the bottom of the screen. Ore transitions between states — minimal, compact, full, active, and listening. Command palette access via keyboard shortcut. Ore is the primary way users interact with the OS.
12 App Spaces
Features are accessed by tapping nodes on the Mycelium, asking Ore, or using the command palette. Each Space opens as a glass-based environment that blooms from the Mycelium. No sidebar — navigation is spatial and ambient.
The 12 Spaces
Briefing
AI-generated morning briefing and daily operational view
Agents
AI agent inventory, configuration, and conversation history
Canvas
Dashboards, reports, forms, and document artifacts
Studio
Visual workflow builder with templates and automation
Governance
Approvals, compliance frameworks, evidence, and audit trail
Analytics
Cost analysis, friction scoring, and operational diagnostics
Connections
Integration marketplace and data flow management
Graph
Knowledge graph visualisation and digital twin
Inbox
Notifications, pending actions, and approval queue
System
Settings, team management, billing, API keys, and security
Profile
Personal energy tracking, habits, and Wheel of Life
Help
Documentation, guided tutorials, and support
The Governance Engine
A bio-inspired 5-layer immune system protects every action. Governance is not a gate at one point — it is a field across the entire stack.
Skin
Rate Limiting & Input Validation
DDoS protection, input sanitisation, format validation. The first line of defence at the network edge.
Innate Immunity
Authentication & Identity
JWT verification, API key validation, SSO/SAML, tenant isolation. Verifies who you are before processing anything.
Adaptive Immunity
Intent Classification
AI-powered analysis of every action. What is this actor trying to do? Anomaly detection and prompt injection checks run here.
Governance
The 7 Laws
Policy enforcement with seven immutable laws: Evidence, Policy, Budget, Approval, Trust, Audit, and Containment. Violations are absolute barriers.
Memory
Trust Scoring & Pattern Learning
Trust scores flow upward from evidence. The system learns which actors, actions, and patterns are trustworthy — and which need closer scrutiny.
Dual-Brain Evaluation
Two models, one verdict
For high-risk actions, two independent AI models evaluate the request separately. Both must agree before the action proceeds. If they disagree, the action is escalated to a human for approval. This eliminates single-model blind spots and provides verifiable consensus.
The 7 Laws
Immutable governance principles
Human-in-the-Loop Action Tiers
Not every action needs approval. The system classifies actions by impact.
AUTO
Immediate execution, no approval needed
Read data, search, generate reports, answer questions
SOFT
Execute with a 30-second undo window
Send notifications, update records, create tickets
HARD
Requires explicit human approval before execution
Send emails, delete data, deploy code, financial transactions
Evidence Chain
Hash-linked, tamper-evident, auditor-ready
Every action in the system — whether executed by a human, an AI agent, or an automated workflow — produces a cryptographically signed evidence artifact. Each artifact is hash-chained to the previous one, creating a tamper-evident audit trail. The chain records who acted, what was decided, which model was used, the governance verdict, and the outcome. Compliance teams can export the full chain for any time period, mapped to regulatory frameworks including SOC 2, GDPR, and the EU AI Act.
Integration Architecture
Connect your entire tool ecosystem. ARQERA is the intelligence layer that sits above your existing stack.
360+ Pre-built Connectors
Connect Slack, Notion, Jira, Salesforce, HubSpot, Xero, and hundreds more via OAuth. One-click setup from the Connections Space. Powered by Nango.
OAuth & One-Click Setup
No API keys to manage. Authenticate with your existing credentials. ARQERA handles token refresh, scope management, and connection health monitoring.
Real-Time Webhooks
Receive instant notifications when events happen in connected tools. HMAC-signed payloads ensure integrity. Filter by event type and route to workflows.
REST API & SDK
Full REST API with 800+ endpoints. OpenAPI specification. Versioned. Rate-limited. Tenant-scoped. Interactive API explorer available in-app.
MCP Server
Connect your AI assistant (Claude, ChatGPT, or custom) to ARQERA via the Model Context Protocol. Your assistant gains governed access to all capabilities.
Custom Integration SDK
Build your own connectors with the ARQERA SDK. Full TypeScript and Python libraries. Webhook registration, event handling, and governance integration included.
Multi-Model AI Strategy
Decision-making is expensive. Execution is cheap. They should not be treated the same. ARQERA separates thinking from doing.
The Core Insight
Most AI systems “think” every time and then immediately act. This is wasteful and dangerous. ARQERA separates the governance brain (which evaluates intent and risk) from the execution engine (which carries out the work). The governance brain runs on premium models and is always on. The execution engine routes to the cheapest model that meets quality requirements for each task.
Governance Brain
Always on
Premium AI models (dual-brain consensus)
Evaluates every action for intent, risk, and policy compliance. Two models must agree on high-risk decisions.
Execution — Budget
80% of tasks
Cost-optimised multimodal models
Routine operations: data extraction, report generation, chart creation, bulk processing. Handles text, image, audio, and video.
Execution — Balanced
15% of tasks
Mid-tier reasoning models
Multi-step analysis, summarisation, classification, and moderate reasoning tasks.
Execution — Premium
5% of tasks
Frontier reasoning models
Complex reasoning, governance verdicts, security analysis, and nuanced judgement calls.
Intelligent Routing
The router selects the optimal model for each task based on cost, quality, and latency requirements. Switch models without code changes.
Platform-Managed Routing
All AI providers managed by ARQERA. Physarum-based routing selects the optimal provider for each task. No API keys to manage.
Custom Model Training
Fine-tune models on your organisation's data. Proprietary adapters trained with consent-gated data. Safety alignment is always enforced.
Platform Stack
A modern, cloud-native stack designed for reliability, security, and global edge performance.
Frontend
Next.js
React
TypeScript
Tailwind CSS
Backend
Python
FastAPI
PostgreSQL
Redis
Infrastructure
Kubernetes
Cloudflare (edge)
Multi-cloud
CI/CD
Observability
PostHog (analytics)
Sentry (errors)
Grafana (metrics)
Evidence chain
800+
API Endpoints
196+
Backend Services
360+
Integrations
9
AI Providers
The 5-Layer Stack
Layer 5
Ambient AI
Ore + Ara across every surface: web, mobile, Slack, Teams, email, API, MCP
Layer 4
Capabilities
12 App Spaces grouped into four bundles: Comply, Operate, Optimise, Managed
Layer 3
Governance
5-layer immune system, 7 Laws, dual-brain evaluation, evidence chain
Layer 2
AI Engine
Governance brain (premium) + execution engine (budget-tier). Thinking and doing separated.
Layer 1
Infrastructure
Cloud-native Kubernetes, edge CDN, managed database, multi-region
Available Everywhere
One operating system, every surface. Ore meets you where you work.
Web (PWA)
Live
iOS
Coming soon
Android
Coming soon
Slack
Live
Microsoft Teams
Live
Live
REST API
Live
MCP Server
Live
Ready to build on ARQERA?
Create your free account and explore the AI-native operating system.