// service
Agentic Network Operations — Intent-to-Fabric
Plain-language network intent becomes verified, on-device configuration changes with a closed telemetry loop — and zero unverified success reports.
shape
8–12 weeks (PoC 4–6 weeks)
Cisco (SONiC · IOS-XR · NX-OS) · Nokia (SR Linux · SR OS) · Juniper · loadbalancer.org (ADC) · NVIDIA · Red Hat (Ansible/RHEL)
outcome
You stop pushing change windows through hand-crafted CLI and start declaring the fabric you want. Multi-agent intent tiers translate business and engineer language into validated changes, Kubernetes-native controllers reconcile them onto live EVPN/VXLAN fabrics (SONiC and Nokia SR Linux), and gNMI telemetry closes the loop so nothing is reported done until the device proves it. Drift is detected and repaired automatically, which converts audit anxiety into a demonstrable verification trail.
what you get
- Fabric intent model and multi-agent orchestration design (AGNTCY/LangGraph patterns)
- Reference implementation reconciling declarative state onto a containerlab-plus-Kubernetes EVPN/VXLAN fabric
- gNMI closed-loop verification and drift-repair pipeline with a published verification policy
- Demo recording and convergence metrics on your representative topology
evidence — public, checkable
- agentic-netops (github.com/mairp) — 4/4 constructs converge; 0 unverified success reports; 6½-min demo
- agentic-netops-srl — the Nokia SR Linux variant
- srl-sros-telemetry-lab — interactive streaming telemetry
- IEEE paper — decentralized zero-touch provisioning of leaf-spine EVPN data centers
The reference build runs a 2-spine / 2-leaf / 4-client SONiC EVPN/VXLAN fabric in containerlab next to a kind cluster. A LangGraph supervisor with mapper, allocator and deployer agents (AGNTCY patterns, A2A/SLIM) turns requests like “stretch this mac-vrf across both leaves” into declarative resources; Go controllers reconcile them with server-side dry-run and deterministic apply; per-node verification gates mark services Ready only after the device confirms. Converged services are re-verified every five minutes — drift is repaired, and said so.
Adoption is staged: start with one construct class on a lab fabric, expand at your pace. Every phase defines its verification bar up front, and you see the evidence on your own fabric before sign-off.
pairs with
Agent-Fleet Platform — Memory, Routing, Observability
A governed multi-agent platform with shared memory, model routing, and full token/cost/trace telemetry — so fleets are operable, not just demoed.
Agentic CloudOps & AIOps — Signal to Action
Your telemetry stops being a wall of alerts and starts closing loops — events correlated, remediated under guardrails, and verified resolved.
Critical-Network Security & Telemetry Assessment
Your utility, railway, ISP or data-center fabric gets a quantum-safe segmentation and streaming-telemetry readiness review — hands-on, not a checklist.