VNClagoon AI

An AI factory you own — technically, legally, operationally.

Two spheres, one platform: the Agentic Suite assists every user interaction, and the Agent Factory orchestrates specialist agents that plan, develop, test, and deploy software — under human authority, always.

The Agentic Suite

Sovereign AI, in every interaction.

One AI Agent across the whole suite — assisting inline in each app, or reasoning across all of them at once. The user picks the sovereignty level per query; nothing leaves your infrastructure without permission.

The VNClagoon AI Agent in standalone mode, querying four connected apps in parallel
Standalone mode — one prompt runs across every connected app in parallel (chat, mail, calendar, projects), with a switchable sovereignty level and an auditable skills library.
The AI Agent panel inside VNCtalk running a local on-device model
Inline in VNCtalk — summarize, draft, translate, and extract tasks in the flow of a conversation. Here it runs fully on-device (Local AI · Llama 3.2 7B via Intel OpenVINO); no data leaves the machine.
Three Levels of AI Sovereignty

The user chooses, per query.

Local AI

Where it runs
On the user's device — CPU, NPU, or GPU
Model class
Llama-class 7B and beyond
Network
none · air-gapped

Private AI

Where it runs
Your GPU cluster — vLLM with Kubernetes autoscaling
Model class
70B-class and beyond, RAG-enabled
Network
within perimeter

Public AI

Where it runs
External frontier models — strictly opt-in
Model class
GPT, Mistral & more
Network
user-authorized

// The user selects the sovereignty level per query. Data never moves without permission.

The Sovereign AI Engine

Enterprise AI, on your hardware.

High-throughput inference

vLLM serves large models on your GPUs; Kubernetes autoscaling keeps latency flat as enterprise load grows.

Sovereign memory & RAG

Every prompt is grounded through an embedding pipeline over pgvector or Apache Solr — context-aware answers from your data, on your hardware.

MCP tool registry

Agents reach the suite and your ERP, HR, and document systems through the open Model Context Protocol — governed, logged, permissioned.

AI managed as code

Model configs, prompt templates, and tool definitions live in Git and deploy through ArgoCD — the AI factory is as reproducible as the infrastructure.

Observable intelligence

Token throughput, model latency, and GPU utilization on the same Grafana dashboards as the rest of the platform.

Auditable by default

Every agent action is logged, traceable, and reversible. Compliance evidence is a byproduct of operation, not a project.

The Agent Factory

Software delivery, orchestrated — with human gates.

The VNClagoon Orchestrator governs specialist agent teams through a four-stage pipeline — with a mandatory human gate at every transition, notified natively through VNCtalk, VNCproject, and VNCmail.

Stage 01

Specify

Agents analyze requirements and draft structured specifications.

⏸ human gate

Stage 02

Develop

Code is generated, reviewed, and refined into tested modules.

⏸ human gate

Stage 03

Test

Automated QA, regression, and security checks with signed reports.

⏸ human gate

Stage 04

Deploy

Release pipelines execute with a complete audit log.

⏸ human gate

// Autonomous speed. Human authority. No exceptions.

Bring sovereign AI to your stack.

See the AI factory running on your own infrastructure — with your engineers, not ours, in control.

Talk to an engineerExplore the platform