Getting started with Totebot
Go from sign up to a live AI agent on your website in an afternoon. A step by step guide through setup, knowledge, and channels, with the tips that save you a week of trial and error.
Totebot turns what your business knows into an AI agent that talks to customers wherever they reach you: your website, WhatsApp, Telegram, email, and more. It works the same whether you run an online store, a rental business, or a busy support inbox. This guide takes you from an empty account to a working agent that answers real questions, and points out the choices that matter along the way.
You can finish every step below in an afternoon. No technical skills are required.
The four things you set up
Everything in Totebot is built from four pieces. Once you see how they fit together, the dashboard makes sense quickly.
- A team is your workspace. It holds your plan, your agents, and your teammates.
- An agent is the AI itself. It has instructions, a personality, and a knowledge base.
- A knowledge base is what the agent knows. You fill it with documents, pages, Q&A pairs, and product data.
- A channel is where customers talk to the agent: your website chat, WhatsApp, Telegram, email, and whatever we add next.
The rest of this post follows that order, because that is the order that gets you to value fastest.
Tip: You do not need a paid plan to start. The Free plan gives you one agent and enough messages to build, preview, and prove the agent works before you spend anything.
Step 1: Create your account and first agent
After you sign up and confirm your email, Totebot walks you through a short setup wizard. It only asks for the decisions that matter on day one:
- What kind of agent you need. Customer support, sales assistant, or both. If your store is already connected, Totebot suggests the right fit. Each choice starts the agent with proven instructions and abilities, so you are not staring at a blank page.
- Which AI model it runs on. A sensible default is preselected. Keep it for now; you can switch at any time, and our companion guide on choosing a model covers when it pays to change.
That is it. Totebot sets the agent up and drops you straight into a preview where you can talk to it. Everything the wizard chose for you is editable later in the agent's settings, including the instructions themselves.
Tip: Pick the agent type that matches what your customers actually ask for today, not what you hope to sell them next quarter. You will tune the agent after you read real conversations, and a grounded starting point gets you there faster.
Step 2: Give the agent something to know
An agent with no knowledge can be friendly, but it cannot be useful. The knowledge base is where Totebot earns its keep.
There are four source types, and you can mix them freely:
| Type | Best for |
|---|---|
| Document | PDFs, Word files, and text files you already have |
| Q&A | Specific, high traffic questions with exact answers |
| Link | Pages you want Totebot to read, like FAQs or policies |
| Text | Quick notes written directly in the dashboard |
Start small and specific. Upload your shipping policy, your returns policy, and your ten most common support questions as Q&A pairs. That alone covers a large share of real conversations.
After you add a source, Totebot takes a short moment to learn it. The dashboard shows you when each source is ready, and if something goes wrong with one, you can simply retry it.
If you run an online store, connect Shopify or WooCommerce here too. Totebot syncs your catalog so the agent can find products, answer questions about options and pricing, and look up order status. Live pricing and stock always come from your store directly, so they stay accurate.
Tip: Every plan includes a knowledge allowance shared across everything you add. More is not better. A focused base of accurate, well organized content beats a sprawling one full of duplicates. If a source has never helped answer a question, remove it.
Step 3: Preview before you publish
Before the agent touches a real customer, talk to it yourself. The live preview runs the exact chat your visitors will see.
Ask it the questions you already know the answers to:
- A shipping question your policy covers
- A product question your catalog can answer
- A question you deliberately did not give it knowledge for
That third question matters most. You want to see how the agent behaves when it does not know something. A good agent says so and offers a next step. If it invents an answer, that is a signal to tighten your instructions and add the missing knowledge.
Tip: Keep a short list of these test questions in a note. Every time you change the instructions, the model, or the knowledge, run the same list. It turns vague impressions into a repeatable check, and you will catch problems immediately.
Step 4: Put it where your customers are
When the answers look right, connect a channel.
Most teams start with the website chat. Totebot gives you a small snippet to add to your site: paste it yourself, drop it into your tag manager, or send it to whoever manages your website. It is a two minute job, and the chat bubble appears on your pages without slowing them down.
If your customers live in messaging apps, connect WhatsApp or Telegram from the channels screen, and email handling is available too. The same agent, with the same knowledge, serves every channel you connect, including the ones we add in the future.
And if you would like a hand at any point, with setting up the agent, the knowledge, the actions it can take, or squeezing more out of one that is already live, message me directly at info@totebot.ai. I read these myself.
Tip: Roll out narrow before you roll out wide. Put the chat on one lower traffic page first, watch a day of real conversations, fix the worst gap, then expand. A staged rollout turns your first customers into a feedback loop instead of an audience for early mistakes.
Step 5: Read what happens next
The work is not finished when the agent goes live. It starts there.
Open the conversations view after the first day. Read real conversations end to end. You are looking for three things:
- Questions the agent answered well, which confirm your knowledge is working
- Questions it answered poorly, which point to gaps to fill
- Questions it should have handed to a person, which tell you where a human belongs
When a conversation needs a person, the agent can hand it over. On paid plans, you can turn on live chat so your team claims and answers those conversations in real time. The handover is smooth: the customer keeps the same thread, and your team picks up with the full history in front of them.
A sensible first week
You do not need to do everything at once. A schedule that works:
- Day one: do the full setup: run the wizard, add your core policies and answers, connect your store or other data, and test it yourself.
- Day two: test internally with your team. Have everyone ask the questions customers really ask, and note what falls short.
- Day three: reach out to us if you want help with optimization, setup, or any question at all. Message me at info@totebot.ai and we will look at your agent together.
- Day four: roll out to your customers and start collecting real conversations.
The agents that perform best are not the ones with the most knowledge on day one. They are the ones whose owners read conversations and close gaps every week.
Where to go from here
Once your first agent is live and answering well, two questions usually come up: which model should it run on, and how do you write instructions that hold up under pressure. Both have their own posts on this blog. Start with the agent that works, then make it better.
Build your own agent
Go from sign up to a live agent in an afternoon.
Keep reading
6 ways teams use the Totebot Help Desk
The AI handles most conversations on its own. The Help Desk is for the ones it should not finish alone. Here are six situations where a shared team inbox earns its place, and exactly what happens in each.
Deflection vs revenue: support is your most underrated sales channel
Most teams grade AI support by tickets deflected. That metric hides the bigger opportunity: every support question is a buying signal. Here is how to treat support as a sales channel.