> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rovax.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Creating an Agent from Scratch

> Step-by-step guide to manually creating an agent in Timely.ai: choosing the model, defining the name, role, prompt, and saving with full control from the start.

Manual creation gives you direct control over every agent parameter. Instead of going through a guided conversational flow, you fill in the identity fields, choose the language model, and write the initial prompt directly in the platform's forms.

This path is recommended when you already have clarity about what the agent needs to do — for example, when replicating a configuration that already works in another workspace, when creating agents with very specific instructions, or when configuration speed matters more than step-by-step guidance.

Any field can be adjusted after creation. There is no permanent commitment when saving — the agent stays in **Draft** status until you publish it to a channel.

## How to create

<Steps>
  <Step title="Go to the Agents section">
    In the Timely.ai sidebar menu, click **Agents**. Click the **+ New agent** button in the upper corner of the page.
  </Step>

  <Step title="Select manual creation">
    In the choice dialog that appears, click **Create manually**. You will be taken directly to the configuration form, bypassing the conversational flow.
  </Step>

  <Step title="Set the internal name">
    In the **Name** field, write the agent's internal identifier. This name appears to your team in logs, reports, and the agent list — it is not shown to the end customer. Use something descriptive like "L1 Technical Support" or "Inbound Sales BR".
  </Step>

  <Step title="Define the agent's objective">
    In the **Objective** field, describe in one or two sentences what the agent should do. This text feeds the initial prompt automatically generated by the platform as a starting point.

    Examples:

    * "Handle customers in post-sale, respond about order status, and initiate exchange processes."
    * "Qualify inbound leads, collect name, email, and budget before passing to the sales team."
  </Step>

  <Step title="Configure the basic options">
    Three behavior options are available in the creation form:

    | Option            | What it does                                                            |
    | ----------------- | ----------------------------------------------------------------------- |
    | Transfer to human | Allows the agent to forward conversations to an attendant               |
    | Use emojis        | The agent can include emojis in responses                               |
    | Split responses   | Long messages are broken into smaller blocks (recommended for WhatsApp) |
  </Step>

  <Step title="Create the agent">
    Click **Create agent**. The agent is created with **Draft** status and appears in the agent list. From here you can access all the advanced configuration tabs.
  </Step>
</Steps>

## Next steps after creation

With the agent created, configure the details in the available tabs:

<CardGroup cols={2}>
  <Card title="Model and parameters" href="/en/guides/agents/settings/model">
    Choose the provider, the LLM model, and adjust temperature and effort for the use case.
  </Card>

  <Card title="System prompt" href="/en/guides/agents/settings/prompt">
    Write or refine the prompt with the advanced editor, version history, and AI improvement.
  </Card>

  <Card title="Knowledge Base" href="/en/guides/knowledge/linking-agent">
    Connect documents, websites, and Q\&A to the agent's knowledge base.
  </Card>

  <Card title="Internal chat" href="/en/guides/agents/settings/internal-chat">
    Test the agent before publishing it to any external channel.
  </Card>
</CardGroup>