Introduction
Agents help administrators and learners complete learning and knowledge related tasks through AI-powered conversational experiences. You can use pre-built agents for common workflows or create custom agents to support your organization’s specific needs.
AgentHub is the central page for creating, configuring, and managing agents. Superadmins can access it from the Agents card in the Launch pad or from the Artificial intelligence control panel.
When available to the user, agents can be selected from the agent menu in the lateral chat panel. Which agents the user can access depends on the user profile, agent’s configuration and trigger mode. For example regarding trigger, proactive agents, such as the Reinforcement agent, run automatically and cannot be started from the menu.
To make agents available to users, enable at least one AI assistant capability for the relevant users in the AI assistant configuration settings of the Artificial intelligence control panel.
Select the History button in the lateral chat panel to access previous conversations with agents.
In addition, when agents access underlying course training materials or assets on behalf of the user, the interacting user is counted as an active user for the billing period. For details on triggering criteria, refer to the Definition of Active user in the Usage dashboard.
Agent usage consumes credits according to the AI credit system overview.
Pre-built agents
Pre-built agents are ready-to-use agents designed for common learning workflows.
From AgentHub, Superadmins can enable or disable each agent and configure its icon, color, title, and description.
Beyond these base settings, pre-built agents’ instructions and capabilities are predefined by Docebo and cannot be adjusted.
Authoring agent
Start from a document, a knowledge source, or a simple prompt. The agent asks a few questions about your audience, depth, and tone, then generates a structured lesson with flashcards, accordions, assessments, and other interactions built in. After the draft, continue refining the lesson with the agent in Creator.
The Authoring agent is selected by default when you open the lateral chat panel in Creator and Content center.
Learn more about the authoring agent.
Marketplace agent
Skip the manual catalog browsing. Describe what you're looking for and get curated recommendations from Content marketplace in seconds. Compare your options, save shortlists for review, and import directly into Course management. When courses in your catalog are flagged for retirement, the agent finds replacements before you have to.
This Marketplace agent is selected by default when you open the lateral chat panel on the Marketplace page.
The Marketplace agent is available to Superadmins on platforms with a Go1 Content Marketplace subscription and using the new Go1 experience introduced in the July 2026 release.
Learn more about the marketplace agent.
Reinforcement agent
When a learner completes a supported course, the agent schedules a session and sends an in-platform notification when they're ready for reinforcement. Each session is personalized to the completed course and adapts based on the learner's previous interactions, so reinforcement becomes more targeted over time. Administrators enable and configure the reinforcement per course.
Learn more about the reinforcement agent.
Custom agents
Custom agents let Superadmins create AI-powered conversational experiences tailored to specific learning and knowledge workflows, such as onboarding, compliance follow-up, learner support, content maintenance, or reporting.
When creating or editing a custom agent in AgentHub, you can configure the following settings:
- Icon, color, title, and description: Define the agent's basic metadata, visual identity, display name, and purpose summary.
- Description: Provide a short overview explaining what the agent helps users accomplish.
- AI instructions: Specify rules for how the agent behaves, priorities it should follow, and operational boundaries.
- Trigger: Set up interaction options for manual execution (custom welcome message and conversation starters), with scheduled runs and platform event triggers coming in future releases.
- Audience: Restrict access to specific users by selecting target Groups or Branches.
- Required level: Restrict access based on administrative level (All users, Power users and Superadmins, or Superadmins only).
- Knowledge: Provide reference materials by uploading files or attaching knowledge collections and learning content, including courses, learning plans, training materials, and assets. For the supported file types and the maximum size of each file, see Supported content types for text extraction and transcript generation.
- Actions: Enable actionable capabilities from a suite of supported Docebo MCP tools (opens in a new tab), with third-party MCP tools support coming in future releases. By default, the agent asks for user confirmation before executing an action. To run an action automatically without confirmation, click the gear icon (Configure) on an action tag and toggle on Auto run.
Your changes are saved automatically as you build your custom agent, so you never lose your work. Your agent stays in draft and is not visible to users until you publish it. Once published, eligible users can access the agent from the lateral chat panel.
To check how an agent behaves before publishing, select Preview in the agent editor. The lateral chat panel opens with the agent marked as Preview, so you can chat with it and test your AI instructions, knowledge, and actions. Update the configuration and preview again until the agent responds as you expect, then select Publish changes.