Create Your First Agent

Agents are the AI assistants that handle your customer conversations in Lyro. This guide walks you through building one from scratch - writing its prompt, picking a model, connecting knowledge and skills, testing it, and publishing it live.

Step 1: Create the agent

From the dashboard, open Agents and click New Agent. Give it a clear name (for example, "Support Agent" or "Sales Helper") - you can rename it any time by clicking the title at the top of the editor.

New agents start as a Draft. Nothing reaches your customers until you publish, so you can configure everything safely first.

Tip: A short, specific name makes agents easy to tell apart once you run more than one.

Step 2: Set the prompt and model

The General tab is where you shape how the agent behaves.

• Description - A one-line summary of what the agent does. The router uses this to decide which agent should handle a conversation.
• Model - The model that powers the agent. Models are grouped by tier so you can balance speed against capability.
• Role - What the agent is permitted to do and what it refuses. Manage roles in Settings.
• Tone - The communication style: Friendly, Professional, Technical, or Casual.
• Visitor context - Contact properties the agent can see. Name and email are always sent; add custom properties on top.
• System prompt - Custom instructions appended to every conversation. Markdown is supported, so you can layer extra context on top of the role.

Choosing a model

Start with a Balanced model. You can change it later and re-test before publishing.

Step 3: Attach knowledge

Open the Knowledge tab and toggle on the bases this agent is allowed to search. If no base is selected, the agent has no knowledge to draw on and will not search at all.

You also choose a knowledge mode:

• Full - Use the knowledge base when relevant, fall back to general knowledge otherwise.
• Strict - Answer only from the knowledge base, and refuse when the answer is not there.

Need to add content first? See train Lyro on your content and the knowledge reference.

Step 4: Add skills

Skills bundle instructions and tools the agent can call. Open the Skills tab and click New skill to create one. Each skill belongs to this agent, so you create, edit, and remove them right here.

A skill can be set to Always on, and any tools it includes show up as a count on the skill card.

Save the agent first if you just created it - skills are added after the initial save.

Escalation

In the Escalation tab, turn on the procedures that decide where the agent hands off when it cannot resolve a conversation - to a human, a ticket queue, or another channel. See set up escalation to your team to configure these.

Step 5: Test in the playground

Before going live, try the agent in the playground. Send realistic questions and watch how it answers, which knowledge it pulls, and when it escalates. Adjust the prompt, model, or knowledge and test again until the responses feel right.

Step 6: Publish

When you are happy, click Publish. Lyro shows a diff of everything that changed since the last published version, and you can add a short note describing the update before confirming.

• Saving creates or updates a Draft without affecting live conversations.
• Publish promotes the draft to a new version and makes it the live config.
• Use Discard draft to throw away unpublished changes and return to the live version.
• The clock icon opens version history, where you can review past versions.

Once published, flip the Active switch in the header to start serving customers.

What's next

• Connect the agent to where your customers are in channels.
• Watch live conversations under conversations.
• See the full agents reference for every setting in detail.