Creating an Agent
- How to navigate to the Agents page and create a new agent
- What each field on the agent creation form does
- How the form changes based on the assistant type you select
Creating an AI agent in AutoTalk lets you automate customer conversations. Each agent has its own name, type, and configuration, so you can run separate agents for sales, support, or scheduling.
Step 1 — Open the Agents page
- In the AutoTalk sidebar, go to Automations.
- Click Agents.
- You will see the agents list page with the title "Agents" and a table of any existing agents.
Step 2 — Start creating a new agent
Click the + (add) button at the top of the list. This opens the agent creation form.
Step 3 — Fill in the top-level fields
The creation form starts with two fields on the General tab, regardless of which agent type you choose:
| Field | Description |
|---|---|
| Name | A text input where you enter a descriptive name for your agent (e.g., "Customer Support Bot" or "Appointment Scheduler"). |
| Assistant type | A dropdown with three options: llm, typebot, or actor. This determines how the agent processes conversations. The actor type is a rule-based assistant that reacts to each message by running an ordered list of Step rules (action sequences) instead of calling an LLM. |
When you choose llm, a Platform type dropdown appears next to Assistant type so you can pick the LLM provider.
Step 4 — Configure type-specific settings
After selecting the assistant type, the form adapts to show fields relevant to that type:
- If you select "llm": Additional fields and five configuration tabs appear (General, Messages, Actions, Tools, and Advanced). See Configuring LLM Models and System Messages and Tools for full details.
- If you select "typebot": Fields appear for connecting your Typebot chatbot flow. See Typebot.
The agent form also has an Advanced tab. The Log level dropdown — which controls logging verbosity for this agent — lives there, inside the Options group (payload path options.log_level); it is not a top-level field. Higher verbosity captures more detail, which is useful during setup and troubleshooting.
Step 5 — Save the agent
Once all fields are configured, click Save. The agent appears in your agents list and is ready to be assigned to a channel.
Assigning an agent to a channel
After creating the agent, connect it to one or more messaging channels so it can respond to incoming customer messages:
- Go to your Channels settings.
- Select the channel you want to automate (e.g., a WhatsApp number or webchat widget).
- In the channel configuration, find the agent or bot assignment option.
- Select your newly created agent.
- Save the channel settings.
The agent will now automatically respond to messages received through that channel.
Choose the "llm" type for flexible AI conversations. You can refine the settings, add tools, and expand system messages after the agent is saved.