2. Your first AI employee
Create your first AI employee
In this article, we will walk through how to create your first employee on the TozheBot platform in just a few minutes, using a website customer support agent as an example.
Account registration
Before you begin, you need to register on the platform and verify your email:
You will then be prompted to create a project by specifying its name, description, field of activity, and website:
You are now ready to create your first agent.
Creating the first AI employee
After completing registration and entering your project details, you will be redirected to the page for creating your first AI-powered employee.
You can also navigate to the employee creation section yourself.
An AI employee is an agent that incorporates everything necessary for operation: a set of instructions, a knowledge base, tools and MCP servers, as well as connectors (external connection sources for AI agents). Below, we will examine each step of creating such an employee.
Selecting a template
To get started, you need to select a template for the employee. The template includes pre-configured instructions for agents, as well as a curated set of tools and connectors:
Let’s look at creating an agent using the example of a technical support specialist.
You can create an unlimited number of staff members and isolate them within specific projects.
Basic agent settings
The next step in creating an AI employee involves configuring the basic settings for the AI agent.
Using the example of a technical support specialist, let’s look at the key configuration parameters (data based on the English-language interface):
- Employee Name - the name of your AI employee.
- Visibility - an agent's scope of visibility. A private agent can be used only within the platform (in isolation) and is best suited for internal tasks (such as processing internal documents or employee onboarding), whereas a public agent can be connected to external interfaces (for example, as a customer support widget on a website).
- Role - a description of the agent's role. For example, "Technical Support Specialist." Essentially, this is your AI employee's job title.
- Tone of Voice - this parameter specifies the tone to use when communicating with the client—for example, being more professional or empathetic.
- Answer Lanuage - the language the agent uses to reply to the user. If automatic mode is selected, the agent will reply in the language the user used to contact it.
- System prompt - these are the core system instructions for your AI agent. They outline the rules it will follow when handling inquiries.
By specifying all the parameters, you provide the agent with instructions and a distinct personality. You can then proceed to connect knowledge bases.
Connecting knowledge bases
The next step is connecting knowledge bases. Knowledge bases represent the experience your employee relies on. You can either create new knowledge bases or select an existing group of knowledge bases. In this case, let's create a new knowledge base using a website:
The agent's knowledge base can consist of a website (automatically ingested from the site's content), a specific resource link, a file (PDF, Word, MD, or TXT format), plain text, product feeds (Yandex, Google), or other types of knowledge sources.
Multiple agents can share overlapping or common knowledge bases, operate without them, or maintain their own exclusive knowledge bases.
After connecting your first knowledge base, you will be redirected to the connector selection page.
Connectors
Connectors - are external sources through which requests can be sent to the agent. Examples include a website widget, a Telegram bot, VK community messages, etc.
You can select existing (previously created) connectors or connect new ones. To configure the connectors after creating the agent, you can navigate to the relevant sections in the control panel.
The final step is selecting the tools.
Connecting tools (agent skills)
The final step in creating an employee is selecting their skills. The skills catalog includes both internal tools (such as the ability to transfer a conversation to a human operator if the agent cannot answer a question) and external ones (such as the Notion MCP server, which allows searching your external knowledge base):
Once the tools are connected, your agent will not only possess knowledge but also learn to perform actions.
The final step involves verifying all parameters and confirming the creation of the agent.
What's next?
Once the employee has been created, you will be provided with useful links - for configuring connectors or verifying your agent's operation.
Before you begin, we recommend testing your agent's performance in the sandbox:
Try writing to your agent and see how they respond to your questions.
We wish you success in bringing your ideas to life!