Skip to main content

​
What Is an Agent?

An Agent is an AI assistant configured to address a specific need or use case within your organization. Each Agent combines a set of parameters that determine its behavior, capabilities, and scope of action:
  • Instruction (system prompt): defines the Agent’s behavior, tone, and guidelines.
  • Model (LLM): the language model used to generate responses.
  • Tools and MCP: the capabilities enabled for the Agent (e.g., Search Documents, Search Visual, MCP Web Search Linkup).
  • Documents and Workspaces: the document sources the Agent can access.
  • Groups / Teams: the users authorized to access this Agent — a key element for data governance.

​
Multi-Tool Reasoning (Agent Mode)

Each Agent is equipped with Agent Mode (V2), which provides multi-tool reasoning capabilities:
  • In-depth analysis: the Agent breaks your request down into sub-tasks.
  • Chained execution: it can autonomously use 1 to 5 consecutive tools.
  • Enriched result: it combines the outputs from these various tools to build a precise, sourced, and comprehensive response.
Important note regarding history: Conversations from version V1 are archived in read-only mode. The new V2 architecture, capable of using multiple tools simultaneously, requires starting a new session.

​
Multi-Agent: What Changes with Wise Wolf

Before Wise Wolf, each organization had a single default Agent, configurable by an administrator. This single Agent defined the instruction, model, and tools for all users. With the Multi-Agent feature introduced in Wise Wolf, it is now possible to associate multiple Agents with a single organization. The goal: specialize assistants for recurring tasks and share them across the enterprise, while maintaining fine-grained control over data access.

​
Selecting an Agent

When creating a new conversation session, a dropdown menu in the interface header allows you to choose an Agent. This menu displays:
  1. “My Assistant” — the organization’s default Agent, always in first position.
  2. Your favorite Agents and those associated with your team, sorted alphabetically (up to 5 displayed).
  3. A “See all assistants” link that opens the full discovery page.
By default, “My Assistant” is pre-selected. Once the first message is sent, the Agent is locked for the entire duration of the session — it is not possible to switch Agents mid-conversation. The active Agent’s name remains visible in the header. Hovering over it displays a tooltip with its details (tools, documents, model, team).

​
Browsing Available Agents

The “All assistants” page displays all the Agents in your organization that you have access to. For each Agent: its icon, name, and a short description. From this page, you can:
  • View an Agent’s details in a modal window (name, description, available tools, accessible documents, model used, associated team).
  • Add or remove an Agent from your favorites by clicking the star icon.
  • Start a conversation directly with the selected Agent.

​
Managing Your Favorites

The favorites system lets you customize your quick-access list in the dropdown menu:
  • Click the star icon from the discovery page or the detail modal to add or remove a favorite.
  • You can only favorite Agents that you have access to through your groups.
  • The dropdown menu displays a maximum of 5 Agents (in addition to “My Assistant”), sorted alphabetically.

​
Creating a Personal Agent

Any user can create their own Agents from the Paradigm interface. These personal Agents are currently not shareable with other collaborators — this capability will be added in a future version.

​
Using Agents via the API

It is possible to specify the Agent to use via a dedicated API endpoint, allowing you to integrate Agent selection into your workflows and third-party applications.

​
Frequently Asked Questions

Can I switch Agents mid-conversation? No. The Agent is locked once the first message is sent to ensure session consistency. To use a different Agent, start a new conversation. Are the Agents I create visible to my colleagues? Not at this time. Agents created from the user interface are personal. To share an Agent with a team, contact an administrator. How can I find out which documents an Agent can access? Open the Agent’s detail modal (from the dropdown menu or the discovery page): the accessible documents and workspaces are listed there. How many Agents can I add to my favorites? There is no limit to the number of favorites. However, the quick-access dropdown displays a maximum of 5 Agents (in addition to “My Assistant”).