Platform · Deployment and Integrations

Deploys where you need it. Integrates with what you already run.

Nexoraa is offered in SaaS, private cloud, on-premise, and air-gapped configurations. Integration is treated as first-class engineering: pre-built connectors for dominant enterprise systems, a connector framework for everything else, and operating patterns that fit existing architectural standards.

Deployment Models

Deployment models.

SaaS (multi-tenant)

Hosted by Nexoraa on a major cloud provider. Tenant isolation is enforced at the platform layer; no shared workloads. Suitable for non-regulated workloads or pilots.

Single-tenant SaaS

Dedicated tenant infrastructure managed by Nexoraa. Common for enterprises wanting managed operations with stricter isolation than multi-tenant.

Private cloud (customer-managed)

Deployed inside the customer's cloud account (AWS, Azure, GCP). Customer retains full data control, network control, and identity boundaries. Operations can be customer-led or co-managed.

On-premise

Deployed inside the customer's data centre. Designed for environments with strict data residency or regulatory restrictions.

Air-gapped

Deployed inside a network without internet egress. LLM models are self-hosted; updates are delivered via signed offline packages.

Reference Architecture

Reference architecture, logical.

The reference architecture is layered. Identity sits at the top, enforcing access on every downstream layer. The Control Plane carries policy, agent definitions, and workflow definitions. The Execution Plane runs the agents and workflows, with Haluvance enforcement as a cross-cutting gate. The Knowledge Layer holds memory, RAG indexes, and structured business knowledge. The Integration Layer carries connectors to enterprise systems. Observability runs alongside, capturing traces, metrics, and audit events from every layer.

Availability

High availability, scale, and recovery.

High availability

Active-active or active-passive clustering with automatic failover; no single point of failure for application servers, databases, vector stores, or message queues.

Auto-scaling

Horizontal scaling on CPU, memory, queue depth, and token throughput. Inference, API, and worker tiers scale independently.

Disaster recovery

Defined RPO and RTO; cross-region or cross-AZ replication; documented and rehearsed DR runbooks.

Backups

Automated daily backups with point-in-time recovery for databases and vector stores. Backup restore tested on a regular cadence.

Capacity planning

Documented capacity model: max concurrent users, tokens per second throughput, storage growth, cost per 1000 requests.

Multi-region

Multi-region deployment supported for latency and data residency. Per-region data control is enforced.

Integration Catalogue

Integration catalogue.

Enterprise systems of record

SAP, Oracle ERP, Workday, Microsoft Dynamics 365, NetSuite, Salesforce, ServiceNow.

Industry-specific systems

Banking core (Finacle, T24, Mambu); capital markets (Murex, Calypso); insurance (Guidewire); healthcare (Epic, Cerner, Facets, HealthEdge); manufacturing (Siemens MES, AVEVA, Maximo); pharma (Veeva Vault, Argus, Trackwise).

Document and content

SharePoint, OneDrive, Google Drive, Box, Dropbox, S3, Azure Blob, on-premise file shares, Confluence, Notion.

Communication and collaboration

Microsoft Teams, Slack, Outlook, Gmail, ServiceNow Now Mobile, Zendesk, Freshdesk.

Identity and security

Azure AD, Okta, Google Workspace, OneLogin, Ping, plus SAML/OIDC-conformant providers; HashiCorp Vault, AWS Secrets Manager.

Observability

Datadog, Splunk, ELK, Grafana, Prometheus, OpenTelemetry endpoints.

LLM providers

OpenAI, Azure OpenAI, Anthropic, Google Vertex, AWS Bedrock; self-hosted Llama, Mistral, Qwen, custom models. Model routing supports per-workload provider selection.

Connector framework

Where a pre-built connector does not exist, the connector framework is the same one used internally. New connectors are written, tested, and shipped without forking the platform.

Developer Surface

API and developer surface.

Nexoraa exposes a versioned REST API surface for workflow execution, agent management, and observability. Webhooks support outbound event delivery with retry, dead-letter, and signature verification. SDKs in Python, JavaScript, and Java are published. An internal event bus enables loose coupling between components and event-driven integration with downstream consumers.

Operating Model

Operating model.

Customer-managed

Customer operates the deployment with Nexoraa engineering support on demand. Suitable for mature platform engineering teams.

Co-managed

Nexoraa operates the platform; the customer retains data and policy control. Suitable for enterprises wanting managed operations with retained governance.

Nexoraa-managed (SaaS)

Nexoraa operates the platform end-to-end. Suitable for environments where data is non-regulated or operating model is centralised.

Next Step

Bring your reference architecture. We will fit Nexoraa into it.