Defining AI Connection Objects

This documentation uses terms and concepts that are common in the generative AI ecosystem. For definitions of terms such as LLM, MCP Server, and tools, see Key Concepts: AI Integration in Automic Automation.

Connection objects contain the parameters that make the communication between Automic Automation and external systems possible. As an administrator, you create the Connection objects that developers and object designers use to configure executable objects.

AI Connection objects act as central infrastructure bridges between Automic Automation and the Automation.AI component, linking AI Jobs to Large Language Models (LLMs), Model Context Protocol (MCP) Servers and tools. By defining an AI Connection object, administrators establish a secure 'sandbox' of approved AI models and tools. This allows object designers to focus on automation logic without needing to manage API keys or low-level protocol configurations.

Unlike other Connection objects (such as SQL or R3), AI Connection objects do not store authentication credentials or API keys. These are managed securely within the Automation.AI component to ensure sensitive tokens are never exposed in tAWI or in Transport Cases.

Benefits of AI Connection Objects

• Centralized management

  Change the underlying model in one place without modifying hundreds of individual Jobs.

• Access control

  By creating different AI Connection objects (for example, one for Finance AI with CSV tools, one for IT Ops AIwith Automic Automation tools), you can control which tools and models are available to different teams.

The Connection Lifecycle

The AI Connection object does not communicate with the LLM directly. Instead, it interacts with the Automation.AI component, which acts as a gateway.

• At Configuration: The Connection object queries the Automation.AI REST API to dynamically populate the list of available models and MCP toolsets.

• At Execution: When an AI Job runs, it references the Connection object. The Automation Engine sends the Job's requirements to Automation.AI, which then handles the secure authentication and the execution of MCP tools (like database queries or system commands) on behalf of the AI.

Prerequisites

• The Automation.AI component must be installed and configured.

• The lists of available LLMs and MCP Servers must be configured in the Automation.AI component.

• Granular security (Least Privilege): Instead of giving an AI model full access to a database, the AI Connection object allows you to expose only specific MCP tools. This ensures the AI can only 'see' the data points required for its specific task.

• Token masking: "Because credentials are stored in the Automation.AI component and not the Connection object, sensitive API tokens are never exposed in logs, AWI exports, or Transport Cases.

This section describes how to define AI Connection objects in AWI. You can also use the Automation Engine REST API to create, read, update and delete AI objects. See https://docs.automic.com/documentation/webhelp/english/all/components/DOCU/26.0/REST%20API/Automation.Engine/index.html?overrideUrls=../Automation.Engine/swagger.json,../Analytics/swagger.json,../Infrastructure.Manager/swagger.json.

An AI Connection object definition consists of several pages:

• Standard pages, available for all object types:

• Defining the General Page
• Defining the Version Management Page
• Defining the Documentation Page

• The AI Connection page

Defining an AI Connection involves selecting a pre-configured model and assigning the necessary toolsets. You do this on the AI Connection page:

1. Select the Model (the LLM) to which you want to establish the connection. The LLMs available for selection are dynamically provided by the Automation.AI component, for example Gemini, GPT-4o and so forth. Selecting a model here determines the 'intelligence' and cost-profile of all Jobs using this connection.

2. In Tools, select the Model Context Protocol (MCP) Servers and specific tools that will be available to the AI Jobs that use this Connection object. This list is populated via the Automation.AI REST API and reflects the MCPs configured within the Automation.AI component. When you select MCP Servers and tools, you are defining the maximum 'capability set' for this connection.

   As an administrator, you use this section to define the "scope" of available capabilities by exposing specific MCPs and tools to the Connection object. Subsequently, when developers and object designers define an AI Job, they can choose from this pre-approved list. The AI Job definition allows for further granular selection, ensuring that only the necessary MCPs and tools from the Connection object are used during execution.

• Automic Automation's Generative AI Capabilities

• Automation Assistant: Streamlining the Administrator's Work

• Defining AI Jobs

• Use Case: AI-Powered Incident Analysis and Resolution with Automic Automation

• Use Case: MCP-Powered Criteria Evaluation