What intoCHAT is and how an agent works
Learn what an intoCHAT agent does on your website: it answers visitors from your own content, collects leads and calls your API, and how the parts fit.
intoCHAT lets you build an AI chat agent that answers visitors on your website from your own content. This page explains what an agent can do and how the parts of the dashboard fit together.
What an agent does
- Answers questions from your content. You give it pages of your website, documents, text snippets and Q&A pairs. For each question it looks up the most relevant passages and writes an answer from them, in the language the visitor writes in.
- Says when it doesn't know. If your content doesn't cover a question, the agent says what it can't confirm instead of inventing an answer.
- Collects leads. A form inside the chat can ask for an email address or phone number. Contact details a visitor types into the chat can be saved as well.
- Takes actions. It can call your own HTTP API during a chat, run a JavaScript function on your page, search the products of a connected store, or show a button that opens one of your pages.
The agent runs on your website: as a chat bubble, inside a page as an iframe, or on its own page behind a direct link. It doesn't hand conversations over to a person. When it can't help, it can collect contact details so your team can follow up.
How the pieces fit together
Every agent has eight tabs, in the order you usually need them:
| Step | Tab | What you do there |
|---|---|---|
| 1. Set up | Setup | Name the agent and write its instructions and guardrails: role, tone and limits. |
| 2. Train | Knowledge | Add website pages, files, text and Q&A pairs, then press Train agent. |
| 3. Style | Appearance | Set theme, colors, font, avatar, launcher, welcome messages and suggested questions. |
| 4. Extend | Actions | Connect your API, functions on your page or your online store, or add link buttons. |
| 5. Test | Playground | Chat with the agent before it goes live. |
| 6. Install | Share | Copy the script tag, iframe or direct link, or use the WordPress plugin or Shopify snippet. |
| 7. Follow up | Leads | See, filter and export the contact details visitors left. |
| 8. Review | Conversations | Read transcripts and reply ratings to see what visitors ask and where answers fall short. |
The Dashboard page shows conversations, messages and leads for the last 7, 30 or 90 days.
Once the script is on your site, you don't need to touch the code again. New knowledge, instructions and appearance settings reach your visitors as soon as you save or train them.
Terms used in these docs
- Agent: one chatbot with its own instructions, knowledge, look and installation code. Your plan sets how many you can create.
- Source: one item of knowledge, such as a crawled page, a file, a text snippet or a Q&A pair.
- Training: intoCHAT splits your sources into passages and indexes them for search. It doesn't change the AI model itself. See How answers work.
- Message: one visitor message that the agent answers. Messages count toward your plan's monthly allowance, including messages you send in the Playground.
- Model: answers are written by one OpenAI model that intoCHAT selects and manages. There is no setting for the model or its temperature.
Where to go next
- Quick start: from sign-up to a live chat widget in five steps.
- Crawl your website: the fastest way to give your agent knowledge.
- Write instructions: set the role, tone and limits.
- Plans and limits: messages, agents, characters and crawls per plan.