Agentic Reasoning, Meet Deterministic Execution: What’s New in Itential Platform 6.5

Getting an AI model to recommend a network change is easy. Letting it take that action on production infrastructure, with the guardrails, approvals, and audit trail your operation actually requires, is the hard part. Itential Platform 6.5 is built around closing that gap. FlowAI, the agentic harness of the Itential Platform, is now generally available. Work Center gives operators one place to handle every human-in-the-loop decision. And Itential Gateway 5.5 pulls external secrets management into the execution path. Reasoning that moves at machine speed. Execution that stays under your control.

FlowAI: The Agentic Harness, Now Generally Available

FlowAI connects AI reasoning to deterministic, governed infrastructure automation. It gives teams a way to design, deploy, and monitor agents that reason through complex goals and take action through the automation assets you already run, producing structured, auditable outcomes without giving up operational control.

That last part is the point. Plenty of tools can put an agent in front of your infrastructure. Very few let it act inside the same RBAC, approval, and audit model that governs everything else you run.

FlowAI is built from a set of surfaces, each mapped to the person who uses it:

• FlowAgent Builder is where builders define an agent’s purpose, instructions, LLM model, and the specific tools it’s allowed to call. Tools can be Platform workflows, adapter and integration APIs, Itential Gateway services, Lifecycle Manager actions, Compliance Plans, and even other agents. Builder also lets you control who can run each agent individually.
• FlowAgent Projects are organizational containers that group related agents and their tools – every agent belongs to exactly one project. Projects add a layer of control on top of agent-level permissions: they enforce a security boundary (agents can only invoke tools you explicitly authorize, and every tool association is auditable), project-level roles (Owner, Editor, and Viewer) that govern who can view, edit, and manage agents, and portability (export a project to move agent configurations between environments, such as from development to production).
• FlowAgent Sessions give operators a chronological execution trace for every agent run: reasoning steps, tool calls, and linked jobs, with the ability to pause, resume, and cancel a session from one view.
• FlowMCP Gateway connects external Model Context Protocol (MCP) servers to the platform, automatically discovers the tools they expose, and surfaces them in Gateway Manager as native Gateway services. Those tools then become available to agents in FlowAgent Builder alongside workflows, APIs, Lifecycle Manager actions, and Compliance Plans, under the same governance model.
• Model Registry lets LLM Admins govern which third-party models are available for reasoning, per environment, with control over which builder groups can access them.

What Actually Changes for Teams Adopting AI

The hard question every infrastructure team is asking right now isn’t whether AI can help. It’s how to let it act without handing it the keys.

FlowAI answers that at build time. An agent can only invoke the tools you explicitly authorize when you create it. Those tools are scoped in FlowAgent Builder, locked at design time, and validated at runtime. Every action is logged. Every tool call is attributed. Agents operate under the same role-based access control model as the rest of the platform, which means an agent never gets a broader reach than the person or group who owns it.

The result is agentic automation you can actually put in front of production infrastructure, because the controls were never optional.

Why This Matters for Agentic Operations

Agentic operations only works when AI reasoning connects to deterministic execution. FlowAI makes that connection native.

Agents and workflows aren’t separate worlds. An agent can call a workflow. A workflow can hand a decision to an agent through the new Run Agent task in Studio, then resume with the agent’s output. Agents can call other agents as tools. Deterministic and agentic automation coexist in a single execution flow, so you can adopt AI incrementally instead of rebuilding what already works.

Underneath all of it, Itential Gateway and Gateway Manager handle tool execution at runtime. Every agent action flows through the same proven infrastructure that powers the rest of your automation. FlowMCP Gateway extends that further, connecting external MCP servers and surfacing their tools natively inside the platform, under the same governance model.

AI adds the reasoning. Itential adds the guardrails.

Explore the documentation on FlowAI →

Work Center: Every Human-in-the-Loop Decision, in One Queue

Not every step in an orchestration should run unattended. Some need a person to review, approve, or supply input before the work continues. Until now, finding those pending tasks meant hunting through Operations Manager. And once agents entered the picture, that problem was only going to get worse.

Work Center fixes it. It gives operators a single queue for every workflow and FlowAgent session waiting on human input. Filter by assigned, unassigned, or all. Sort by how long a task has been waiting, so the work that matters most rises to the top.

Open a task and you get the full context to act: a work item header, a horizontal timeline of every manual task in the job (completed and active), and the current decision front and center. Completed steps show who handled them and when. Finish a task, and if the workflow produces the next one, the view updates on its own. No trip back to the queue. Work Center also introduces Quick Form, a new manual task type that renders any JSON array as an interactive table for structured, per-row operator input.

One detail worth calling out: Work Center is included with Platform 6.5 for every customer. A FlowAI purchase is not required. Whether or not you’re running agents yet, human-in-the-loop workflows finally have a real home.

Explore the documentation on Work Center →

What Else Shipped in Platform 6.5 & Itential Gateway 5.5

Platform 6.5 and Itential Gateway 5.5 also deliver a set of capabilities aimed at execution flexibility, security architecture, and operational integration.

Platform 6.5

Gateway Manager DB Import & Export

Gateway Manager 1.1.1 exposes Itential Gateway 5 database import and export as API-callable methods, available through the Itential Platform API, workflow tasks, or the Gateway Manager UI. Teams can trigger configuration imports and exports programmatically, without embedding iagctl in their pipelines. Both JSON and YAML are supported, and Git-sourced imports are available through the Platform API.

Async Task Support

The Workflow Engine now supports a new async task type. Long-running external systems can call back to the Workflow Engine on completion, instead of forcing a workflow to block or poll while it waits. Cleaner orchestration for the integrations that don’t finish on your schedule.

Compliance Plan Events

Configuration Manager now emits start and complete events for Compliance Plans, consistent with Lifecycle Manager Resource Actions. External systems can trigger workflows off Compliance Plan execution directly, without relying on job events.

Cisco NSO 6.7 Support

Platform 6.5 adds compatibility with Cisco NSO 6.7, including support for new commit parameters and updated version detection. One migration note: teams moving to NSO 6.7 that use doProvision with the legacy flags-based options format will need to update to the new structured object format. Details are in the release notes.

Read the Itential Platform 6.5 release notes →

Itential Gateway 5.5

External Secrets

Itential Gateway 5.5 retrieves credentials at runtime from external secrets management systems, and never stores the resolved value in the Gateway database or returns it to the platform. Supported providers include HashiCorp Vault (KV Secrets Engine v2), CyberArk CCP, and custom plugins. The new $GatewaySecret variable reference lets teams use externally managed credentials across Inventory Manager device credentials, Run Code tasks, Python service decorators, and Gateway-executed integrations. Secrets stay where they live.

FlowAI Tool Execution

Itential Gateway 5.5 adds support for FlowAI tool execution, letting agents invoke Gateway services as tools directly from FlowAgent Builder. This is the runtime path that turns an agent’s decision into action on real infrastructure.

Read the Itential Gateway 5.5 release notes →

From First Agent to Full Production

The headline of Platform 6.5 is FlowAI. But the shape of the release says something bigger about where the platform is going.

FlowAI brings agents into production under governance. Work Center gives the humans in the loop a real place to work. External Secrets meets enterprise security teams where their credentials already live. Every piece assumes the same thing: that AI and people operate the same infrastructure, through the same controls, with the same audit trail behind every action.

That’s what agentic infrastructure operations looks like when it’s built for production, not a demo. From your first agent to full production operations, governed by default.

Platform 6.5, Itential Gateway 5.5, and Gateway Manager 1.1.1 are generally available as of July 1, 2026. FlowAI requires a separate purchase and Platform 6.5 or later. For the full technical release notes and upgrade guide, visit docs.itential.com. Questions? Reach out to your CSM or [email protected].