AI Agents

Agent refers to the artificial intelligence assistant through which your users interact with. You create an agent from a template and configure it to have specific characteristics and knowledge base and give it appropriate tools and deploy it either through a web widget or a phone number.

What you can do

• Create agents from templates tuned for specific use cases or begin from a blank custom configuration.
• Define the agent's name, first message, base prompt, supported languages, and attached knowledge base.
• Enable built-in System tools or connect your own callable tools to extend the agent's capabilities.
• Configure voice behaviour through Voice Playbooks, with control over the speech-to-text, text-to-speech, and realtime models, as well as voice selection, language, and pace.
• Test the agent in a live side-by-side preview before publishing.
• Deploy to an embeddable chat widget or assign a telephone number for inbound calls.
• Monitor performance through dashboard metrics covering active sessions, top intents, and a fallback report. Review every past conversation in History.

How it works

When you create an agent, Enok saves it as a draft with the template's personality, your greeting, and a default chat widget configuration. Nothing is live at this stage; all fields can be edited freely.

1. Personality and knowledge: Your prompt and first message define the agent's behaviour and tone. When a knowledge base is attached, the agent performs semantic retrieval across your indexed articles, pulling the most relevant passages to ground responses rather than generating answers from scratch.
2. Tools: Each enabled tool becomes an action the agent can take mid-conversation when the tool's description matches what the situation calls for, looking something up, transferring a call, writing a record to an external system.
3. Voice and chat: For voice agents, a Voice Playbook determines whether speech is handled through a staged speech-to-text-to-LLM-to-speech pipeline, a realtime TTS model, or a fully realtime end-to-end model. For chat agents, a separate chat model powers typed replies. See Configure an agent for both.
4. Deployment. On deployment, the agent's status moves to Queued while infrastructure provisions, then transitions to Live. Enok notifies you when the process completes.
5. Conversations and insights. Every session, whether ongoing, completed, or closed by the user and closed on idle is recorded with a full transcript. These sessions populate the dashboard's metrics, intent analysis, and fallback report.

Building an agent is a short loop: create it here, then configure its personality, tools, and models, deploy it as a widget or phone line, and watch the dashboard and history once people start talking to it.

Create an agent

1. Open AI Agents from the sidebar.
2. Click New Agent. If no agents exist, a Create new Agent button appears in the empty state.
3. Enter an Agent name.
4. Under Choose a template, review the available options. Each shows a short description of its default behaviour. Select the closest match, or pick Custom for a blank slate.
5. Click Create. The agent opens to its detail view, ready to configure.

Your new agent opens straight to its detail view, ready to configure.

Next, head to Configure an agent to give it a voice, knowledge, and tools.

Manage your agents

All agents come listed under AI Agents with their types, statuses, and creation and modification timestamps.

Change the name of an agent: Launch the agent and go to Configure → General, modify the name in the Name field and save changes. The new name will reflect wherever the agent is mentioned, even in ongoing chats.

Remove an agent: Click the three dots "..." menu next to the agent's row in the list and choose Delete. The deletion process can also be initiated from the agent's own Actions menu.