3.7. Agent setup

Agent setup

You can fine-tune your agents in a separate section of the control panel:

In the list itself, you will see the active agents in the current project, their visibility, and the number of active knowledge bases and tools.

Upon opening the settings, you will have access to broader options for customizing the agent's personality, rules, tools, and limits.

General settings

General settings include the agent's name and description, their photo (used in widgets), and the agent's visibility (public or private).

Ruleset

In the rules section, you specify instructions for your agent. When creating AI employees from a template, these rules are applied automatically, but you can customize them as you see fit.

You can configure the agent's role (job title), tone of voice, response language, additional instructions, and system prompt. All these instructions are stored in the database in encrypted form. You can provide instructions in any language (though English or Russian is recommended), just as you can generate responses to users.

Agent skills

This is a major section that transforms your agent from a mere conversational partner into a functional business unit. Here, you can configure rules for handing off to a human operator (if necessary) and integrate knowledge base groups, platform tools, and external MCP servers from the catalog. We do not recommend creating a "jack-of-all-trades" agent; instead, it is better to create separate agents for specific roles. Just as in real life, a specialized agent will handle a task more effectively and safely than a generalist.

Additional

This is the final configuration section, where you can set limits on the number of conversations and configure automated follow-ups for your agent. Automated follow-ups allow the agent to send a message to the user based on predefined rules if there is no response.

If you need help setting up agents - you can contact us.