Connect
The 400 systems that don't talk start talking — and your AI can act on all of them, not just read.
200+ connectors·CDC & streaming·MCP-native·Reverse actions
One platform. Five surfaces. Zero silos.
Most "AI platforms" are bundles — five products sharing a logo and an invoice. SphereIQ is architected the other way: one control plane, one knowledge substrate, one governance layer that every request passes through — with five pillar applications on top.
How is SphereIQ architected?
SphereIQ is one platform with five pillar applications — Connect, Company Brain, Enterprise Twin, AI Factory, and Governance — sharing a common control plane (identity, tenancy, billing, telemetry). Governance is enforced as a horizontal layer on every request, and the Twin's entity graph is the shared reference model for the whole system.
Read the platform the way it deploys.
Company Brain
A memory that never resigns: knowledge graph, enterprise RAG with citations, and a semantic layer.
Knowledge graph·Enterprise RAG·Semantic layer·Persistent memory
Enterprise Twin
A living model of systems, processes, and people. You can't automate what you can't see.
System discovery·Dependency graphs·Process mining·AI-readiness scoring
AI Factory
Manufacture a governed AI workforce — with evaluation gates no agent ships without passing.
Agent builder·Workflow orchestration·Evaluation harness·Template library
Governance
Policy in the request path, an immutable audit ledger, and cost attribution to the agent.
Policy engine·Guardrails·EU AI Act & GxP packs·AI Economics Intelligence™
Executive Console
The CIO's single pane of glass: adoption, agent utilization, risk posture, and AI Economics Intelligence™ — cost vs. savings, attributed and defensible.
Governance isn't a pillar bolted on the side. It's the plane everything passes through.
The stack, honestly drawn — six layers, one of which every single request is forced to cross.
Executive ConsoleChat & agent surfacesSlack / Teams / email
policyguardrailsPII maskinghallucination checksapprovalsaudit ledgercost metering
— enforced on every request
model gateway & registryembeddingsagent runtimeevaluationmemory (Engram)
knowledge graphvector storesemantic layerTwin entity graphmetadata
200+ connectorsCDCevent busAPI gatewayMCP hubreverse actions
SSO/SAMLRBACmulti-tenancyobservabilitydeployment manager
Enterprise plumbing included.
The parts procurement asks about on the first call — identity, provisioning, access, and failover — are in the platform, not on the roadmap.
Single sign-on
SSO via SAML and Azure AD, plus Google OAuth — your identity provider stays the front door.
SCIM v2 provisioning
Users provisioned and deprovisioned from your directory automatically. Leavers lose access the day they leave.
API tokens
Programmatic access for your own systems and pipelines, governed like every other request.
LLM failover
Automatic cross-provider failover between OpenAI and Anthropic models — a provider outage is not your outage.
Cost accountability
Every AI call is attributed per feature, per user, and per session.
Chargeback exports by user, team, or department (CSV/JSON) · Budget alerts and enforcement · Token-level carbon accounting with CSRD/ESRS E1 reporting
Every material event lands on one audit ledger.
With tamper-evident, Ed25519-signed receipts that can be verified against a published key. Not a log you trust — a receipt you can check.
Receipt shown as rendered by the platform · values illustrative
Three deployment models. One platform.
SphereIQ Cloud
Multi-tenant SaaS
Fastest onboarding. Full governance plane. The default for mid-market teams that want value in weeks.
SphereIQ Dedicated
Single-tenant VPC
Your own AWS environment, managed by Sphere. The enterprise default — isolation without the ops burden.
SphereIQ Private
Your cloud, your keys
Deployed into your environment via infrastructure-as-code. Built for GxP, financial services, and data-residency mandates.
Architecture questions, answered.
SphereIQ is model-agnostic by design. A model gateway routes across Anthropic (primary), open-weight models, and others via AWS Bedrock or self-hosting — with cost-aware routing, fallbacks, and per-model governance policies. No pillar hard-codes a model.
Model Context Protocol is the open standard for connecting AI to tools and data. Every SphereIQ connector and pillar capability is exposed as an MCP server with governed scopes — so agents built anywhere can use your enterprise safely, and you're never locked to one agent framework.
SOC 2 program in progress, HIPAA-ready configurations, GxP validation documentation for life-sciences deployments, and EU AI Act / NIST AI RMF mapping built into the Governance pillar. Dedicated and Private deployments support customer-held keys.
See the whole platform on your own data.
A working session with a Sphere architect: your systems, your use cases, a live environment.