/
Deploying an AI Agent

Deploying an AI Agent

  • Verified

    • Overview

    Once an AI Agent is fully designed, configured, and tested, it needs to be deployed to be used in your application or environment. Deployment options support different contexts — client-side embedding, server-side processes, chatbot platforms, or integrations via industry protocols.

    This page explains how to launch an Agent into real use, the different deployment modes available, and key considerations like authentication, domain configuration, and embedding styles.

    Steps for Deploying an AI Agent

    After designing and configuring your AI Agent, you can deploy it to your application or website.

    1. First, make sure to save the agent into a subdirectory of the modules directory.

      1. Also ensure that this subdirectory is included in the pathlist of your config.js file.

    2. In the IDE menu, select Launch → Deploy this Agent to open the deploy dialog:

      87a9fb12-270d-4639-b0f9-e25cff7313e1.png

    Table of Contents

    1. The Deploy Agent dialog will appear.

    2. Select the desired options, copy the provided code using the icon in the top-right corner of the code snippet, and follow the instructions for pasting it into your application:

      2025-10-16_13-12-00.png

    Deployment Modes

    Select the mode that best fits your architecture. Each mode may require different embedding code or setup.

    Deploy as Web App Onclick Event

    1. When deploying as a Web App Onclick Event, you must create your own button (or some other interface element like a hyperlink or an icon) to trigger launching the agent.

    2. Then paste the provided code into the appropriate event.

    Deploy as Web App Embed Icon

    1. When deploying as a Web App Embed Icon, paste the provided code into your application.

      1. For example, you might place this code in the onload event of your screen/application.

    2. A “chat” icon to trigger the agent will be rendered in the bottom right corner of the application:

      2025-10-16_13-17-23.png

    Deploy as Server Side Code

    This option is ideal for background or batch processes

    Use this option to interact directly with the agent from backend logic. This capability allows the agent to be deployed more autonomously. You can embed the capabilities of AI directly into your applications and workflow without an interactive chat interface.

    For more details on how to use this option, see the pjs.askAgent() API.

    Deploy as an MCP Server

    MCP (Model Context Protocol) is an AI integration standard that allows external AI applications like Claude Desktop, ChatGPT, and Windows Copilot to directly interface with your Profound AI agents. This deployment option enables users to access your data and systems through their existing AI subscriptions, eliminating separate API costs and providing a standardized way to integrate with major AI platforms.

    For more detailed information on how to configure the server for MCP, see here.

    Deploy as Slack Bot

    To deploy the agent as a Slack Bot, you must have admin rights to your Slack workspace.

    1. Copy the provided Slack App Manifest and click the Create New Slack App link.

    2. On the Slack API site, click Create New App and then select “From an App Manifest”.

    3. After the App is created, you must copy the App Credentials into your config.js file.

    4. Restart the server.

    Deploy as OpenAI Plugin

    Deploying your agent as an OpenAI Plugin allows you to interface with other assistant ecosystems, such as the ChatGPT website or Microsoft Copilot.

    Copy the provided URL that hosts your Agent’s plugin manifest (i.e., the JSON file describing your plugin) and provide it when asked for the OpenAI Manifest URL.

    Deploy on Another Domain

    This option should be used if your application is running on a separate domain or port than the Profound AI server.

    With this option, it’s important to also account for the following:

    1. CORS must be enabled for the Profound AI instance;

      1. See Enable CORS for more information;

    2. Each domain must match in regards to using HTTP vs. HTTPS;

      1. Meaning, if one domain is set up to use HTTPS, the other domain must also be set up to use HTTPS;

    Deploy in Profound UI / Genie Interface

    The minimum Profound UI release that supports Profound AI is Version 6, Fix Pack 24.0.

    Select this option if your application is based on Profound UI or Genie interfaces (this includes Profound.js Rich Display applications). The latest versions of these interfaces include the necessary Profound AI code and do not require an additional <script> tag to be pasted into your HTML code.

    Deployment Details Configuration

    This section describes various additional options available for Agent deployment.

    Create Signing Keys for Profound UI / Profound.js Authentication

    For detailed information on this method of Agent authentication, see the following page: https://profoundlogicsupport.atlassian.net/wiki/x/FwC-WQ.

    This option provides an option for the Profound AI agent (running on IBM i or accessing IBM i via Profound.js Remote Connector) to automatically authenticate based on Profound UI / Genie or Profound.js Rich Display File (RDF) session logins.

    2025-10-16_13-36-06.png

    Alternate Prompt Placeholder

    The default placeholder is “Send a message…”. You can adjust this text when deploying your agent. To change the message, simply type a different placeholder entry and it will be reflected in the provided code snippet.

    Alternate Loader Message

    The default loader message when the AI Agent is processing a request is “Thinking”. You can adjust this text when deploying your agent. To change the message, simply type a different message and it will be reflected in the provided code snippet.