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

# Setting Up Your First Agent

> Create a Template or Personalised BDR agent and send your first autonomous outreach.

This page walks through creating an agent in either mode. If you haven't yet, read [How BDR agents work](/bdr-agents/overview) first to decide between **Template** and **Personalised**.

<Note>
  **Personalised agents should be set up through [Alva](/ai-assistant/talking-to-alva), not built from scratch in the UI.** Jump to [Personalised agents](#personalised-agents) below. Template agents can be set up either way.
</Note>

## What every agent needs

Whatever the mode, an agent needs:

<Steps>
  <Step title="At least one active Discovery profile">
    The agent needs companies to target. Create and activate an [ICP or other discovery profile](/discovery/creating-a-discovery-profile) first, and wait for some results to come through. This includes both company searches and [People Searches](/discovery/people-search).
  </Step>

  <Step title="A connected email account">
    The agent needs an inbox to send from. [Connect your email](/email/connecting-an-email-account) in Settings → Email Connections.

    Leave this unset to run the agent in [simulation mode](/bdr-agents/how-the-agent-works#simulation-mode) — it evaluates companies and logs what it *would* do, but sends nothing. A great way to test.
  </Step>
</Steps>

Then, depending on mode:

* **Template** also needs **at least one [email template](/email/email-templates) group** for the agent to send from.
* **Personalised** also needs a completed **playbook** and a successful **preview** — see [Personalised outreach](/bdr-agents/personalised-outreach).

## Template agents

Template agents send your approved templates verbatim, with merge fields filled in per company and contact. You can create one from **Discovery → AI Agents → Create Agent**, or simply ask Alva.

### Name and instructions

Give your agent a descriptive name (e.g. "Healthcare Agency Prospector") and write **instructions** that guide its decision-making:

* **Who to prioritise** — which roles or personas (e.g. "Focus on CTOs and VP Engineering")
* **When to escalate** — what situations should be flagged for human review

<Note>
  In Template mode, don't put messaging guidance in the instructions — the agent uses your email templates for the words. The instructions are about *who* and *when*, not *what to say*. (Personalised agents handle *what to say* through the [playbook](/bdr-agents/personalised-outreach) instead.)
</Note>

### Target profiles

Select which discovery profiles the agent should work from. The agent takes companies from these profiles and begins outreach.

#### Minimum match score

Optionally set a **minimum match score** (1–100). When set, the agent ignores companies scoring below that threshold — even ones already imported — and works through the rest best-fit first. This is most useful with a [People Search](/discovery/people-search), where contacts are imported *before* their companies are scored for fit. See [How the agent chooses who to work on](/bdr-agents/how-the-agent-works#minimum-match-score).

### Email templates

Choose which **template groups** the agent can use. The agent selects the most appropriate template for each prospect and fills in the merge fields.

<Warning>
  A Template agent with no template groups cannot send any emails. Assign at least one before going live.
</Warning>

### Email connection

Select which connected email account the agent should send from. This determines the sender address and signature. Leave it unset to run in [simulation mode](/bdr-agents/how-the-agent-works#simulation-mode).

### Allowed tools

By default all tools are enabled. You can restrict which actions the agent may take:

| Tool               | What it does                                            | Cost |
| ------------------ | ------------------------------------------------------- | ---- |
| **Send Email**     | Send outreach emails                                    | Free |
| **Create Contact** | Save contacts to your CRM                               | Free |
| **Update Status**  | Update outreach status on a company                     | Free |
| **Escalate**       | Flag a company for human review                         | Free |
| **Draft Reply**    | Draft a reply to a prospect's response, for your review | Free |

### Escalation notifications

Optionally add email addresses to notify when the agent escalates a company for human review. See [Managing replies](/bdr-agents/managing-replies).

## Personalised agents

Personalised agents write original copy from a playbook instead of sending fixed templates — a more powerful but more involved setup.

<Warning>
  **Set up Personalised agents by asking Alva.** Alva is fully briefed on the correct way to configure one: it builds the playbook with you, runs sample **previews** so you can see the copy before anything is sent, and only lets the agent go live once the [activation checklist](/bdr-agents/personalised-outreach#going-live) passes.

  Creating a Personalised agent from scratch in the Agent Creation UI is **not encouraged** — the playbook and preview steps are easy to get wrong by hand. Once the agent exists, **small tweaks** in the UI afterwards (adjusting the budget, hours, or an offer) are perfectly fine.
</Warning>

To get started, open Alva and say something like:

> "Set up a Personalised BDR agent for my healthcare ICP."

Alva will walk you through the [playbook](/bdr-agents/personalised-outreach#the-playbook), show you [previews](/bdr-agents/personalised-outreach#previewing-before-you-launch), and confirm the agent is ready before it goes live. Full details are on the [Personalised outreach](/bdr-agents/personalised-outreach) page.

<Tip>
  If you just say "set up a BDR agent", Alva will ask whether you want **Template** or **Personalised** before doing anything — it won't guess.
</Tip>

## Launching the agent

However you created it, a new agent starts **Paused**. Review its configuration, then click **Start** when you're ready.

When active, the agent processes companies on a regular schedule (controlled by its [run pace](/bdr-agents/configuring-budgets-and-hours#run-pace)). For each company it checks enrichment, chooses a contact, sends outreach, tracks responses, follows up, and escalates when human judgement is needed. See [How the agent chooses who to work on](/bdr-agents/how-the-agent-works) for the ordering.

## What happens after launch

The agent appears on the **Discovery → AI Agents** page with live stats:

* **Emails sent** — total outreach emails dispatched
* **Budget spent** — how much of the monthly budget has been used
* **Status** — active, paused, or completed

Click any agent to open its detail page with the [activity feed](/bdr-agents/reading-the-activity-feed), outreach breakdown, and performance metrics.

<Tip>
  Start with a small budget and a slow run pace (1 company per run). Watch the agent's first few actions, review what it sends, and adjust before ramping up. For Personalised agents, run in [simulation mode](/bdr-agents/how-the-agent-works#simulation-mode) or review previews with Alva first.
</Tip>