When your agent is prompted, it takes into account the knowledge base you provide.

1. Edit the knowledge base

To do this, go to the Agent section in your Clice account and click Knowledge base.

Organize the knowledge base like this:

• On the right, you have the list of categories. Click one to open it.
• When you open a category, you will see all the articles inside. You can easily create, edit, and delete them.

This structure has no impact on your conversational agent. It simply helps you navigate among hundreds of articles.

It is important to follow this logic: 1 article = 1 specific topic (you can have multiple articles per topic, but do not mix several topics in one article).

<aside> <img src="/icons/info-alternate_gray.svg" alt="/icons/info-alternate_gray.svg" width="40px" />

You can create almost as many articles as you want, but no more than 1600 characters per article.

</aside>

<aside> <img src="/icons/info-alternate_gray.svg" alt="/icons/info-alternate_gray.svg" width="40px" />

This is a Chatbot‑oriented knowledge base.

This means your users will never see this knowledge base. You can therefore create tailor‑made articles for the chatbot to read.

For example, in an article feel free to:

• Add keywords to refine your agent’s search
• Give it custom instructions like “if the user asks this, set this tag and reply with that…”
• Write roughly without super user‑friendly sentences if you need to trim to stay under the 1600‑character limit
• But if you just want to copy‑paste your documentation, that also works very well! </aside>

2. Train the agent

After you edit an article, your agent forgets it. You must click the Train button (red, top right, you cannot miss it).

If you have many articles to teach the agent, go back to the main Agent section. You will find an alert telling you your agent is not trained on some articles, and it will offer to train them all at once.

<aside> <img src="/icons/warning_gray.svg" alt="/icons/warning_gray.svg" width="40px" />

Training the agent is not optional!

If you leave an article in an unlearned state, your agent does not know it and, even worse, it still knows the old version if the article was modified after training.

</aside>

<aside> <img src="/icons/checkmark-line_gray.svg" alt="/icons/checkmark-line_gray.svg" width="40px" />

That’s it!

Want to go further than just answering questions with your Agent? See Agentic capacities.

</aside>