By James Aspinwall, co-written by Alfred Pennyworth (my trusted AI) — March 7, 2026, 12:45
Netris automates the network fabric underneath GPU clouds. WorkingAgents governs the AI agents that run on top of them. One makes multi-tenant AI infrastructure possible. The other makes multi-tenant AI agent operations trustworthy. Together, they close the gap between “infrastructure is provisioned” and “agents are operating safely.”
What Netris Does
Netris is the leading Network Automation, Abstraction, and Multi-Tenancy (NAAM) platform, purpose-built for GPU clouds and enterprise AI factories. They posted 622% year-over-year ARR growth in 2025, captured 12% of the global neocloud market in 10 months, and onboarded 15 AI cloud operators across 20+ data centers — every deployment live in production.
Core capabilities:
- Automation — auto-configures VRFs, VXLANs, BGP, and networking elements in under two minutes with closed-loop assurance
- Abstraction — delivers cloud-grade network abstractions (VPCs, peering, elastic IPs, load balancers) via web console and REST API
- Multi-Tenancy — hardware-level tenant isolation across bare metal, VM, and container workloads, enforced at switch and DPU levels
- Multi-Fabric — Ethernet (including NVIDIA Spectrum-X), InfiniBand, NVLink, NVIDIA BlueField DPUs, virtual, and edge networking
Netris was the first ISV validated by NVIDIA for AI Network Automation. Their partnership with vCluster delivers the industry’s first full-stack Kubernetes multi-tenancy for AI infrastructure — hard isolation of tenant clusters spanning physical and virtual nodes.
Their customers are GPU cloud operators, neoclouds, sovereign AI clouds, and enterprises building AI factories. The problem they solve: turning idle GPU capital expense into sustainable revenue by making multi-tenant GPU infrastructure operationally viable.
What WorkingAgents Does
WorkingAgents is the governance and control layer between AI agents and the systems they interact with. Three gateways, one control plane:
- Unified LLM Routing — control which models agents use and how they access them
- Agentic Workflow Control — define, supervise, and enforce how agents take actions
- Enterprise MCP and A2A Tools Access — connect agents to internal tools with least-privilege permissions
Per-user access control, encrypted permission keys, audit trails on every action, 86+ MCP tools (task management, CRM, alarm scheduling, push notifications, system monitoring), and per-user SQLite databases. Agents inherit the user’s permissions — one identity, one set of rules, full accountability.
The Architectural Mirror
Netris and WorkingAgents solve the same problem at different layers:
| Concern | Netris (Network Layer) | WorkingAgents (Agent Layer) |
|---|---|---|
| Multi-tenancy | Hardware-level network isolation | Per-user database and permission isolation |
| Abstraction | VPCs, elastic IPs, load balancers | MCP tools, natural language interface |
| Automation | Auto-configure VRFs/VXLANs/BGP in 2 min | Auto-schedule alarms, tasks, notifications |
| Self-service | REST API + web console for network ops | MCP server + web UI for agent ops |
| Governance | Network policies enforced at switch level | Agent permissions enforced at tool level |
Netris abstracts network complexity so operators don’t need network engineers for every tenant. WorkingAgents abstracts operational complexity so enterprises don’t need custom integration code for every agent. Same pattern: take something that requires specialized expertise, make it self-service with guardrails.
Synergy Areas
1. The Missing Application Layer for GPU Clouds
Netris solves the network layer. Their Kubernetes multi-tenancy partnership with vCluster solves the orchestration layer. But GPU cloud operators still need an application layer — the tools and workflows their tenants’ AI agents actually use.
WorkingAgents fills this gap:
- Tenant onboarding — new GPU cloud customer signs up → Netris provisions isolated network → vCluster creates isolated Kubernetes namespace → WorkingAgents creates per-user database with scoped permissions. Complete tenant isolation from network fabric to agent behavior.
- Operational tooling — GPU cloud tenants get more than compute. They get task management, scheduling, CRM, notifications — all through MCP tools their AI agents can call directly.
- Agent-to-infrastructure bridge — tenant’s AI agent needs to scale GPU allocation → calls WorkingAgents MCP tool → WorkingAgents checks permissions → triggers infrastructure API → Netris handles the network reconfiguration → WorkingAgents logs the action and schedules a cost review.
GPU cloud operators using Netris + WorkingAgents offer a complete platform, not just isolated compute.
2. Network Operations Governed by Agents
Netris auto-configures networks in under two minutes. But who initiates that configuration? Today: human operators via web console or REST API. Tomorrow: AI agents.
WorkingAgents governs that transition:
- A tenant’s AI agent determines it needs a new VPC for a training cluster → calls a WorkingAgents MCP tool wrapping Netris’s REST API → WorkingAgents validates the agent’s permissions (does this tenant have VPC creation rights?) → Netris provisions the network in 2 minutes → WorkingAgents creates a task to track the resource, schedules a teardown alarm for the estimated job completion, and notifies the platform team
- If the agent tries to modify another tenant’s network → WorkingAgents blocks the action at the permission layer, before it ever reaches Netris → audit log records the attempt
Netris enforces isolation at the hardware level. WorkingAgents enforces isolation at the agent intent level. Defense in depth — two independent enforcement points for the same tenant boundary.
3. Neocloud Differentiation — Agent-as-a-Service
Netris captured 12% of the global neocloud market. These neoclouds compete fiercely on price and performance. Differentiation comes from value-add services.
WorkingAgents enables a new service tier:
- Basic tier: Raw GPU compute with Netris network isolation (commodity)
- Platform tier: GPU + Kubernetes multi-tenancy (vCluster) + network automation (Netris)
- Agent tier: GPU + Kubernetes + network + governed AI agent infrastructure (WorkingAgents) — task automation, CRM, scheduling, notifications, MCP tool access, all within the tenant’s isolation boundary
The agent tier is where margins live. Tenants don’t just rent GPUs — they rent a governed AI operating environment. Stickier, higher-value, harder to replicate.
4. Sovereign AI and Compliance — Full Stack Isolation
Netris supports sovereign AI cloud deployments. Sovereign clouds require provable isolation at every layer. Today, Netris proves network isolation. With WorkingAgents:
- Network layer (Netris): Hardware-enforced VRFs, VXLANs, DPU-level isolation. Provable via network topology.
- Compute layer (vCluster): Hard Kubernetes tenant isolation. Provable via cluster boundaries.
- Agent layer (WorkingAgents): Per-user encrypted permissions, per-user databases, audit trails on every agent action. Provable via access control logs.
A sovereign AI cloud auditor can trace isolation from the network switch → through the Kubernetes cluster → to the individual agent action. Three independent layers, three independent audit trails, one coherent compliance story.
5. Closed-Loop Assurance Across the Stack
Netris’s “closed-loop assurance” automatically validates that network configurations match the desired state. WorkingAgents provides the same pattern at the agent layer:
- Netris detects network drift → auto-corrects → logs the event
- WorkingAgents detects an agent exceeding its permissions → blocks the action → logs the event → notifies via Pushover
- Together: a tenant’s agent triggers a network change → Netris configures it → Netris’s closed-loop validates the network state → WorkingAgents’s alarm system validates the operational state (did the training job start? did it complete? did costs stay within budget?)
Closed-loop at the network level. Closed-loop at the operational level. The entire stack is self-healing and self-monitoring.
6. Multi-Fabric Awareness for Agent Workflows
Netris uniquely supports multiple fabric types — Ethernet, InfiniBand, NVLink, DPUs. Different AI workloads need different fabrics:
- Training jobs → InfiniBand for RDMA performance
- Inference serving → Ethernet for cost efficiency
- Multi-node inference → NVLink for GPU-to-GPU bandwidth
WorkingAgents can make fabric-aware decisions:
- Agent receives a training job request → WorkingAgents checks the job type → selects the appropriate fabric via Netris API → provisions resources on InfiniBand → tracks the job → when complete, releases InfiniBand resources and re-provisions on Ethernet for inference serving
- All governed by permissions: this tenant can use InfiniBand, that tenant is restricted to Ethernet. WorkingAgents enforces, Netris provisions.
The Partnership Opportunity
For Netris: WorkingAgents extends their value proposition from “we automate your network” to “we automate the entire operational stack.” Their neocloud customers get agent governance as a platform feature. Sovereign cloud customers get application-layer compliance. The 15 operators Netris onboarded in 2025 are all potential WorkingAgents deployments.
For WorkingAgents: Netris solves a problem we can’t — network-level multi-tenancy. WorkingAgents can govern agent behavior all day, but without hardware-enforced network isolation, tenant boundaries are software-only. Netris provides the foundation that makes our per-user isolation meaningful at the infrastructure level. Their 622% growth trajectory and NVIDIA validation signal a company worth partnering with early.
For the joint customer: A GPU cloud that’s isolated from network switch to agent action. Self-service from VPC provisioning to task scheduling. Automated from BGP configuration to alarm escalation. Compliant from network topology audit to agent permission log.
Concrete Next Steps
- MCP tool integration — Wrap Netris’s REST API as WorkingAgents MCP tools: VPC create/delete, network status, tenant isolation verify, fabric selection. Estimate: 2-3 days for 6-8 tools.
- Joint tenant onboarding demo — New customer signs up → Netris provisions isolated network → WorkingAgents creates governed agent environment → tenant’s first AI agent runs a training job end-to-end with full audit trail.
- Neocloud pilot — Approach one of Netris’s 15 neocloud operators about deploying WorkingAgents as the agent governance layer, validating the “Agent tier” service model.
Netris turns raw network hardware into a multi-tenant GPU cloud in weeks. WorkingAgents turns raw AI agents into governed, auditable employees in hours. Same philosophy — automation, abstraction, multi-tenancy — applied at adjacent layers of the stack. The combination delivers what neither can alone: a GPU cloud where every layer, from the network switch to the agent’s last API call, is isolated, automated, and auditable.