AI Agents

Create custom intelligent assistants

In ENSPACE, you can create custom AI Agents within each workspace. Unlike BENI — which is the default assistant with access to all tools — custom agents allow you to define exactly what each assistant can do, how it behaves, and which tools it uses.

Once created, the agent becomes available in the workspace’s floating AI chat, where you can select which agent you want to use for the conversation. Agents respect the permissions of the member interacting — if you do not have access to a category, the agent will not have access either.

Agents can be used for automations, internal assistants, support, data analysis, document review, and other scenarios.

AI Tools

Each type of agent comes with a set of pre-selected AI tools, optimized for the best performance of that type. The tools define the agent’s capabilities — for example, whether it can create records, query tasks, send emails, or manage members.

For the full list of available tools, see AI Tools.

Agent structure

Creating an agent is divided into three tabs:

Essential tab

Defines the agent’s basic and structural information:

• Agent Name (required) — identification of the agent in the workspace.
• Language Model (required) — selects which AI model will be used to generate responses. The model directly impacts quality, speed, and cost of interactions.

• Agent Type (required) — defines the specialization and the tools the agent will have access to:
  • Chat / Text — focused on chat interactions and text content generation. Suitable for query assistants, content creation, and support.
  • Reviewer — specialized in reviewing documents, identifying errors, and suggesting improvements. Ideal for quality control tasks. In practice, the reasoning model is optimized for critical text analysis.
  • Form (coming soon) — specialized in collecting and structuring data via forms.
  • Multi-task (coming soon) — capable of executing complex sequences of varied actions.
• Status (required) — Active (available for use) or Inactive (disabled).
• Description (optional) — summary of the agent’s purpose.
• Instructions (optional) — general context defining who the agent is and its main objective.

Customization tab

Defines the agent’s detailed behavior:

• Icon — visual identification.
• Agent Role — functional identity (e.g., "Legal Assistant", "Financial Analyst", "Technical Reviewer").
• Agent Approach — reasoning guidelines (e.g., "Always ask for context before responding", "Prioritize objective answers").
• Agent Rules — limits and restrictions (e.g., "Never share sensitive data", "Do not respond to topics outside the scope").
• Writing Style — formatting and tone (e.g., technical, formal, bullet points, concise).

Test tab

Allows you to test the agent before creating it permanently. Enter a message and click Run Test to validate whether the rules, role, and model are delivering the expected result.

The test does not create the agent or make it available in the workspace.

How to create an agent

Manage existing agents

After creation, the agent can be edited, deactivated, or deleted. All changes take effect immediately after saving. Deleting an agent does not remove the data or interactions it has already generated — those records remain in the workspace.