# AI Calling Agent

- **URL:** https://agentline.cloud/build/ai-calling-agent
- **Category:** Outbound
- **Keywords:** AI calling agent, AI call agent, outbound AI calling, AI call assistant, AI calling assistant, AI phone call agent, AI that makes phone calls

> An AI call assistant that places outbound calls, talks, and reports back

## What is an AI calling agent?

An AI calling agent is a voice agent with its own phone number that places outbound calls, holds the conversation, and returns a transcript and outcome to your system. With AgentLine you create the agent with a prompt, get a US or Canada number on a plan from $20 a month, and start each call with one API or MCP request. Calls beyond the plan's 100 minutes are $0.12 a minute.

This is the general outbound AI calling template: confirmations, callbacks, follow-ups and quick check-ins. Give it a goal, the facts it may share, and what it may agree to. It calls, handles the person who answers, leaves a voicemail when nobody does, and you read the transcript to see what was agreed. Use the cold call template for prospecting lists and the appointment reminder template for scheduled confirmations.

See also: [AI cold calling template](https://agentline.cloud/build/cold-call-agent) · [Appointment reminder calls](https://agentline.cloud/build/appointment-reminder) · [Lead follow-up (GTM) agent](https://agentline.cloud/build/gtm-agent) · [Compare AI calling platforms](https://agentline.cloud/compare)

## How it works

1. Create an AgentLine agent with a goal, the facts it may share, and what it may agree to.
2. Give the agent a US or Canada number from your plan.
3. Start one outbound call per person with the API, the MCP server, or your own agent.
4. Read the transcript and outcome, then update your CRM, calendar or task list.

## Greeting

Hi, this is Sam calling on behalf of Northwind. Is now a good time for a quick question?

## Voicemail

Hi, this is Sam from Northwind. I called with a quick question about your request. Call this number back when convenient. Thanks.

## System prompt

```
You are an outbound AI calling agent for {company}. You call people who expect to hear from {company} or asked for a callback.

Goal for this call: {goal}

Facts you may share:
{facts}

You may agree to:
{allowed_outcomes}

Rules:
- Say who you are and why you are calling in the first sentence.
- Keep each spoken turn under 20 words. Ask one question at a time.
- If the person asks to stop being called, confirm and end the call.
- If they want a human, collect the best callback time and end politely.
- Never invent prices, dates or commitments that are not listed above.
- Before hanging up, read back what was agreed.
```

## Python

```python
from agentline_ai import AgentLine

client = AgentLine(api_key="al_live_...")

agent = client.agents.create(
    name="AI Calling Agent",
    voice_id="female-1",
    initial_greeting="Hi, this is Sam calling on behalf of Northwind. Is now a good time for a quick question?",
    voicemail_message="Hi, this is Sam from Northwind. Call this number back when convenient.",
    system_prompt="""You are an outbound AI calling agent for Northwind.
Goal: confirm the customer still wants the Tuesday 10am install.
You may agree to move it to Wednesday 10am. Read back what was agreed.""",
)

client.numbers.buy(agent_id=agent.id, country="US")

call = client.calls.create(agent_id=agent.id, to_number="+14155557890")
print(call)
```

## Node.js

```ts
// No published Node package. REST against https://api.agentline.cloud
const API = "https://api.agentline.cloud";
const headers = {
  Authorization: "Bearer al_live_...",
  "Content-Type": "application/json",
};

const agent = await fetch(API + "/v1/agents", {
  method: "POST",
  headers,
  body: JSON.stringify({
    name: "AI Calling Agent",
    voice_id: "female-1",
    initial_greeting: "Hi, this is Sam calling on behalf of Northwind. Is now a good time for a quick question?",
    system_prompt: `You are an outbound AI calling agent for Northwind.
Goal: confirm the customer still wants the Tuesday 10am install.
You may agree to move it to Wednesday 10am. Read back what was agreed.`,
  }),
}).then((r) => r.json());

await fetch(API + "/v1/numbers", {
  method: "POST",
  headers,
  body: JSON.stringify({ agent_id: agent.id, country: "US" }),
}).then((r) => r.json());

const call = await fetch(API + "/v1/calls", {
  method: "POST",
  headers,
  body: JSON.stringify({ agent_id: agent.id, to_number: "+14155557890" }),
}).then((r) => r.json());
console.log(call);
```

## Tell your AI

```
Create an outbound AI calling agent on AgentLine. Follow https://agentline.cloud/skill.md then apply https://agentline.cloud/build/ai-calling-agent.md
```

## FAQ

**How is an AI calling agent different from a power dialer?**

A power dialer connects a human rep to the next number. An AI calling agent holds the conversation itself, follows your prompt, and returns a transcript. Use it for routine calls and route anything it cannot handle to a person.

**How much does outbound AI calling cost on AgentLine?**

Plans start at $20 a month with one US or Canada number, 100 minutes and 400 SMS included. Calls beyond the plan are $0.12 a minute, billed per second after a one-minute minimum, and a call that never connects is free. For example, 300 minutes on the Individual plan is $20 + 200 × $0.12 = $44.

**Can my existing agent start the calls?**

Yes. Claude Code, Cursor, Codex, OpenClaw, Hermes or your own service can start calls through the MCP server or the REST API and read the transcript when the call ends.

**Is outbound AI calling legal?**

It depends on who you call and where. In the US, calls that use an AI-generated voice are treated as artificial-voice calls under the TCPA, so you generally need prior express consent, and some states add their own rules. Call people who asked to hear from you, honor opt-outs, and check the rules for your use case.