Itential demonstrates building FlowAI network agents in minutes
Itential published a demo showing an engineer build and deploy a FlowAI network agent in minutes to run dynamic PyATS tests, generate interface health reports, and send results to Slack and email, relevant to network operations teams.
Research overview
The vendor demonstration recorded a practitioner creating a FlowAgent with natural-language system and user prompts, attaching a Model Context Protocol (MCP) and selected tools, and launching the agent into production within minutes.
The run completed dynamic tests against device interfaces in under 400 seconds and produced a detailed health report with notifications delivered via Slack and email.
Key findings
The agent executed PyATS-based test generation and execution, evaluated every interface on a device, and reported a summary of healthy and problematic components, including a single medium-priority IPSec tunnel issue.
Notifications and a score-based report were produced automatically and made available in the platform, Slack, and email as shown in the demo.
Technical breakdown
The builder workflow included composing a system prompt and user instructions, selecting only required MCP tools, binding a deterministic project, and choosing an Large Language Model (LLM) provider and model for agent behavior.
The agent used attached PyATS utilities to generate and run tests dynamically, and the platform bound a testbed YAML to enable execution against one or multiple devices without manual script authoring.
Product update
The demo highlighted FlowAI features for assembling an agent toolbelt, selecting MCP tools selectively, and configuring notification targets before starting an on-demand agent mission.
Agent activity logging in the UI showed tool calls and message exchanges, and the platform displayed test outcomes, findings, and distribution of the report to configured channels.
Operational impact
The demonstrated workflow removed the need to hand-write PyATS jobs, Ansible playbooks, Representational State Transfer (REST) integrations, or custom notification code by using natural-language prompts and bound tools to produce executable tests and reports.
The vendor positioned the approach as repeatable for running interface health checks across single or multiple devices and for scheduling or on-demand use by network teams.
Leadership perspective
The presenter, John Capobianco, described building his first agent during a platform preview and referenced his prior experience with PyATS while showing the end-to-end agent creation and execution steps.
He narrated the sequence of configuring prompts, attaching tools and projects, launching the agent, and reviewing the resultant notifications and report within the platform.
This “Blog Signals brief” is a fact-based summary of the vendor blog and highlights the demo's primary operational details for enterprise technical decision-makers.