Institutional Procurement

Predictable Sovereign Infrastructure Investment.

Fixed monthly deployments with 100% intellectual property ownership. Eliminate variable consumption billing, opaque per-token surcharges, and public cloud leakage.

SINGLE-TENANT COMPUTE

Sovereign Model Cluster

Dedicated, air-gapped private model cluster running inside your own VPC or bare-metal environment with zero external telemetry.

$3,570

per month, billed annually

Dedicated private model instance (GCP africa-south1 or on-prem)
Zero data exfiltration guarantee via bwrap & nftables drop-all egress
OpenAI-compatible REST and low-latency gRPC endpoints
100% irrevocable client ownership of fine-tuned weights
Predictable flat monthly fee (zero variable token markups)
99.9% availability SLA with continuous kernel sentinels

Procure Sovereign Cluster

Recommended Architecture

MOST DEPLOYED

Enterprise Neural Mesh

Full institutional intelligence: Sovereign clusters unified with Cortex sub-2.4ms GraphRAG and autonomous digital workforce fleets.

$8,330

per month, billed annually

Everything included in Sovereign Model Cluster
Cortex Knowledge Graph (KùzuDB property graph + SQLite FTS5)
Sub-2.4ms deterministic recall with exact clause & page citations
Up to 500,000 corporate documents continuously indexed
Connectors for Microsoft 365, Google Workspace & SQL databases
5 Autonomous Digital Worker fleets with human approval checkpoints
Immutable SHA-256 Merkle audit trail for every agent action
Priority 4-hour systems architect SLA & dedicated Slack channel

Deploy Neural Mesh

CUSTOM INFRASTRUCTURE

Institutional Multi-Entity Fleet

Custom air-gapped compute topology for multinational financial institutions, legal practices, and sovereign public institutions.

Custom

tailored architectural procurement

Multi-region VPC clustering or on-premise Dell PowerEdge appliances
Hardware-enforced TPM 2.0 silicon attestation & LUKS encryption
Unlimited corporate document ingestion & multi-graph federation
Unlimited autonomous digital workers across departments
Air-gapped voice intelligence & forensic call audit pipelines
Dedicated on-site resident systems engineer & 1-hour critical SLA
Custom fine-tuning & domain adapter engineering on proprietary corpora

Schedule Fleet Architecture Review

Technical Specification

Architecture Feature Comparison

Detailed technical and security capabilities across each deployment tier.

Capability / Attribute	Sovereign Cluster	Enterprise Neural Mesh	Institutional Fleet
Privacy & Infrastructure Sandboxing
Execution Environment	Single-Tenant VPC	Single-Tenant VPC / Hybrid	Bare-Metal Racks / Multi-VPC
Network Egress Filtering	nftables eBPF drop-all	nftables eBPF drop-all	Physical Air-Gap / Hardware Diode
Model Weights Ownership	100% Client Owned	100% Client Owned	100% Client Owned
Hardware Silicon Attestation	Confidential VM	Confidential VM	TPM 2.0 Hardware Root-of-Trust
Cortex Knowledge Engine & GraphRAG
Graph Engine Architecture	Optional Add-on	KùzuDB Property Graph + FTS5	Multi-Graph Distributed KùzuDB
Query Retrieval Latency	< 15ms	< 2.4ms	< 1.5ms Sub-Millisecond
Included Ingested Documents	Up to 50,000	Up to 500,000	Unlimited Corporate Corpora
Verifiable Source Citations	Standard	Exact Page, Line & Clause	Exact Forensic Merkle Hash
Autonomous Workforce & Governance
Autonomous Software Workers	1 Sentinel Agent	5 Coordinated Fleets	Unlimited Custom Fleets
Human-in-the-Loop Checkpoints	Basic	Granular Role-Based Access	Multi-Signature Hardware Approval
Audit Trail & Verification	Standard Logs	SHA-256 Merkle Ledger	Cryptographic Write-Once Archive
Engineering SLA & Support	99.9% / Next-Day	99.95% / 4-Hour Response	99.99% / 1-Hour Dedicated Lead

CFO & General Counsel Assurance

Commercial Procurement Governance

Zero Variable Consumption Billing

Unlike public hyperscaler APIs that charge escalating fees per million tokens, Compas operates on fixed monthly infrastructure allocations. You never receive variable overage surprises.

Intellectual Property & Weight Retention

Under Section 4.2 of our Master Services Terms, all proprietary fine-tuning adapters, graph schemas, and embeddings remain 100% the permanent, confidential intellectual property of your enterprise.

Data Sovereignty & Jurisdictional Lock

Our infrastructure deploys into specified legal jurisdictions (e.g. South Africa POPIA Act 4 of 2013 or EU GDPR 2016/679). Network traffic never traverses unauthorized international boundaries.

Dedicated Systems Architect SLA

Enterprise Neural Mesh and Institutional Fleet deployments include a designated systems architect for ingestion tuning, custom adapter engineering, and continuous operational support.