What an agent is

An agent is the rep who actually makes the calls. Not the one who means to, or gets to it tomorrow, or lets the list rot in a spreadsheet. It dials, it listens, it answers, it asks the next question, and it is just as sharp and just as friendly on the hundredth call as it was on the first. You give it three things:

1. Who to call. The campaign decides this. See What a campaign is.
2. What to say. Its opening line and its instructions. See The prompt.
3. What you want. The goal, like “book a meeting” or “get them on the phone.”

Then it picks up the phone and goes to work, one call at a time.

​

You brief it, you don’t script it

Here’s the part that surprises people. You do not write out the whole conversation. You can’t, because you don’t know what the other person will say. Instead, you brief the agent like a new hire:

• Here’s who you are and who you work for.
• Here’s why you’re calling.
• Here’s what to do if they say yes, and what to do if they say no.
• Here are the lines you must never cross.

The agent fills in the actual back and forth by itself, in the moment. Your job is to give it a clear goal and good rules, not a word for word script.

​

What you decide about an agent

For the controls on your current screen, see Agent Builder controls. Every agent is made of a few simple choices. Each has its own guide:

​

Analysis and actions have different owners

Define what to extract from a transcript in what the agent records. Configure what happens because of that result in what happens after the call. Booking, transferring or tagging while the conversation is still live belongs to in-call capabilities.

​

Agents that run on TFU Live

Some agents carry a blue TFU Live tag instead of a campaign type. They are the same product with a different calling runtime: one machine that hears and speaks directly, so there are no speed, expressiveness, language or background settings and almost everything is written instead. They have a brain and a voice. See TFU Live agents.

​

One agent, many places

• Clone an agent copies a working agent into another sub-account. The copy is yours to change.
• Agent split tests run two to five versions of one agent in one campaign and tell you which books more.
• Master Agent runs one agent across every client sub-account and keeps the copies in step.

​

In the API

An agent record, from Get agent. Ask for the smallest view you need: view is full by default, script returns just the prompt, campaign just the campaign and tools just the installed tools.

• Create agent takes locationId, name, agentType and capabilityIds, plus optional config and campaign, and accepts an Idempotency-Key.
• Update agent takes name and any part of config. Fields you omit are left alone.
• Delete agent removes the agent and its campaign together. It requires cascade=agent_and_campaign so the deletion cannot happen by accident.
• Create test call and Create web agent session let you hear it. See Test your agent.
• Every field above is defined in the Glossary.

TFU Live has a dedicated complete-agent authoring contract and API reference.