AppsFlowAI

FlowAI overview

FlowAI is the agentic layer of Itential Platform, connecting AI reasoning to deterministic, governed infrastructure automation. Use it to design, deploy, and monitor agents that reason through complex goals and take action through Itential’s existing automation assets, with full auditability and the operational controls your team relies on.

Why use FlowAI?

FlowAI closes the gap between AI capabilities and production-grade infrastructure operations.

Governed agentic automation

Agents can only invoke tools you explicitly authorize at design time. Every action is auditable, every tool call is logged, and agents use the same role-based access control model that governs the rest of Itential Platform.

Hybrid execution model

FlowAI lets agentic and deterministic automation coexist in a single execution flow. Agents can call workflows, workflows can invoke agents through a Run Agent task in Studio, and agents can call other agents as tools. You can build composable architectures of specialized agents and adopt AI incrementally without replacing existing automation.

Purpose-built surfaces for each persona

Each role has a dedicated interface with appropriate access controls:

MCP connectivity

FlowMCP Gateway extends Itential Gateway to connect external MCP servers and surface their tools natively in the Platform, bringing external AI ecosystems and automation frameworks under Itential’s governance model.

Core components

FlowAI is made up of several integrated components that span design, administration, and runtime.

Model Registry

Model Registry defines which third-party LLMs are available for agent reasoning. LLM Admins configure one or more Provider Profiles per environment, selecting a provider, enabling specific models, and controlling which builder groups have access. Both on-prem and cloud deployments support bringing your own provider with customer-supplied credentials. Cloud deployments can also use the Itential-managed provider. For more information, see Model Registry.

Agent Projects

Agent Projects are organizational containers that group related agents and their tool associations. Agent projects establish the security boundary for agent tool access and enforce role-based access control at the project level using Owner, Editor, and Viewer roles. Projects are the unit of portability for export and import of agent configurations across environments. For more information, see Manage agent projects.

Agent Builder

Create and configure agents inside an agent project. Each agent has a prompt, input variables, a selected set of tools it can invoke, and an LLM model for reasoning. You control who can run it. Agents are organized into agent projects, which serve as the security boundary and unit of portability across environments. For more information, see Create and run agents.

Agent Sessions

Agent Sessions is the runtime monitoring surface for operators. It provides a chronological trace for each agent run, including reasoning steps, tool calls, and any agent sessions or workflow jobs the agent called. You can pause, resume, and cancel sessions, and navigate linked sessions and jobs from a unified view. For more information, see Monitor agent sessions.

FlowMCP Gateway

An extension of Gateway 5 that connects registered external MCP servers, automatically discovers their tools, and exposes those tools to Itential Platform through Gateway Manager as native Gateway services. MCP-backed tools participate in the same RBAC model and invocation patterns as all other Gateway services. For more information, see FlowMCP Gateway overview.

How FlowAI fits into Itential Platform

FlowAI builds on and interoperates with the rest of Itential Platform rather than replacing it. Agents draw from a broad tool catalog at runtime, including:

  • Platform workflows
  • Adapter and integration APIs
  • Gateway services (which include Python and Ansible-based services)
  • Lifecycle Manager actions
  • Compliance Plans
  • Other agents

Operations Manager is where you create agent triggers, using Agent as an automation type alongside workflows. All existing trigger types (manual, schedule, API, and event) are available for agent automations. Access control is managed at the automation level consistent with how workflow triggers are governed.

Gateway Manager and Itential Gateway handle tool execution at runtime, meaning every agent action flows through the same proven infrastructure that powers all Platform automation.

Itential also provides the Itential MCP Server, which works in the opposite direction, exposing Platform capabilities (including workflows, device management, and Gateway services) to external AI assistants such as Claude and GPT. Where FlowMCP Gateway brings external MCP tools into Platform for agents to use, the Itential MCP Server allows external AI tools to call into Platform directly. For more information, see the Itential MCP Server repository.

FlowAI is available as an RedHat Pa

System requirements

FlowAI requires the following infrastructure versions:

  • Itential Gateway 5.5 or later
  • Gateway Manager 1.1 or later

Gateway handles all tool execution at runtime. Confirm these versions are deployed in your environment before configuring FlowAI.

Get started with FlowAI

To build and run your first agent, follow this sequence:

1

Configure the Model Registry

Set up one or more LLM profiles that define the provider connection, available models, and which groups can access them.

2

Create an agent project

Create a project to organize your agents, control who can build in it, and export or import agent configurations across environments.

3

Build and run an agent

Create agents inside the project. Each agent references an LLM profile and selects the tools it can invoke at runtime.

4

Monitor in Agent Sessions

View running and completed sessions, review traces, and pause or cancel sessions when needed.

Common use cases

Intelligent troubleshooting and remediation - Deploy agents that reason through observed network conditions, evaluate remediation options, and invoke the appropriate workflows, APIs, or services to resolve issues without requiring manual escalation at every decision point.

Adaptive provisioning - Use agents to interpret high-level provisioning requests, determine the correct configuration sequence based on current infrastructure state, and orchestrate the resulting workflows end-to-end.

Multi-agent orchestration - Build ecosystems of specialized agents that collaborate and delegate based on domain expertise. A diagnostic agent can assess infrastructure state and hand off to a remediation agent, which then invokes the right workflows, APIs, or services to run the fix. You can also invoke an agent from a Studio workflow using the Run Agent task, letting the workflow hand off a specific decision to an agent and resume with the agent’s output.

Extending automation with external intelligence - Use FlowMCP Gateway to incorporate tools from external MCP servers, such as AIOps platforms or domain-specific data sources, directly into Platform workflows and agent tool sets.

Demos and resources

See FlowAI in action with hands-on demonstrations: