Closing the Context Gap Between AI Agents and Network Infrastructure

AI agents generate plausible-sounding output about network infrastructure. But without accurate, structured data about what’s deployed and how it connects, that output falls apart in production. Most agent deployments stall not because the models are wrong, but because the context is missing.

Kris Beevers from NetBox Labs will walk through the emerging AI agent ecosystem for network infrastructure, showing how NetBox captures the structured context agents need, and how NetBox Copilot and the NetBox MCP server make that data accessible through natural language and standard protocols.

John Capobianco of Itential will then take it live. Using Itential’s Platform and FlowAI technology, John will build a FlowAgent entirely in natural language — no REST APIs, no Python, no Ansible, no YAML. He’ll selectively bind NetBox MCP tools alongside pyATS to power two live workflows: one that discovers live network state and pushes it into NetBox, and another that reads from NetBox to drive configuration changes on the network. The result is a repeatable, programmable AI agent that keeps your source of truth aligned with reality.

You’ll walk away with:

• How NetBox provides the structured context AI agents depend on
• How NetBox Copilot and the MCP server give agents access to infrastructure data
• How to build a ReAct agent with Itential FlowAI: writing data into NetBox, reading it out for automation
• How Itential FlowAgents deliver repeatable, reusable, programmable automation that keeps your network platform aligned with your source of reality
• What governance controls should be in place before agents touch live infrastructure
• Where to start: small use cases and a path from pilot to production
  

This session is for network engineers, infrastructure teams, and IT leaders exploring how AI agents can accelerate operations without compromising the accuracy and compliance that production environments require.