Knowledge

One screen per brand lists everything your AI can read — knowledge bases, documentation sets, files, website pages and Q&A — with how much each one contributes and which of them visitors reach on your public web address.

Where to find it

Brand → Knowledge. The page heading reads Knowledge sources.

There is no organization-wide list of knowledge bases any more, no organization-wide list of documentation sets, and no Context tab on the AI agent. You see one brand's knowledge at a time; switching brand switches the list.

The content belongs to the brand, not to a chatbot. Whatever sits on this screen is what the brand's AI answers from — in chat, on phone calls, in the inbox's suggested replies and in posts you write with the social composer. There is nothing to attach and nothing to connect. The marketing autopilots (Social Posting's Autopilot and Newsletter Autopilot) may write from it too, but only from the parts ticked for them on the brand's Content sources page.

The top of the screen

A summary card sits above the list, with four things in it:

  • Public help site — your brand's address (for example acme.yaplet.help) with a pill: Live when a knowledge base or documentation set is on it, Nothing on it when the brand has an address but nothing is served there (visitors who open it see nothing), or No address when the brand has none yet — then a Set an address button takes you to Brand settings. Under the address it names what is being served.
  • Sources — how many sources the AI reads, and across how many kinds (for example Across 3 kinds).
  • Indexed — how many searchable pieces the AI has stored from all your sources together. Hover the label for the explanation. On the older free workspaces that keep a limit it also shows 212 of 250 on the free plan.
  • Website last read — when your websites were last checked, with a note such as Checked every hour. If nobody can pay for it at the moment the note says Paused until you add credit or choose a plan. With no website added it shows a dash and No website pages added.

The groups on the screen

Under the card, a Sources heading introduces the groups.

Knowledge bases & documentation

A collapsible group, headed Knowledge bases & documentation on screen. Structured content you write and organize yourself: knowledge bases (articles for customers) and documentation sets (technical pages). One of each can also be served on your public web address.

Everything else your AI reads

The other collapsible groups: Website, Files, Text snippets, Questions & answers and API tools — pasted text, question-and-answer pairs, uploaded files, website pages we have read for you, and API tools. None of these are ever public — they exist only to be read by the AI.

Each group can be collapsed, and a group with more than ten rows also gets a search box and pages.

Adding knowledge

You add to any group with the Add knowledge button at the top of the page. It opens a window that asks what Yaplet should learn from, in two parts:

  • Teach Yaplet — Website (add your site once; Yaplet learns its pages and keeps up with every change), Files (PDF, DOCX or TXT documents you already have), Text snippet (paste any text Yaplet should know, like a policy), Question & answer (a question and the exact answer Yaplet should give) and API tool (let Yaplet look up live data, like an order status).
  • Help content customers can read too — Knowledge base (help articles for your public help site and for Yaplet) and Documentation (technical pages for your public site and for Yaplet).

A card you cannot use stays visible with the reason instead of disappearing: Ask an admin for access to add this. when your team role lacks the right, or Available on a higher plan. for Documentation when your plan does not include it.

What a row shows

Each knowledge base and documentation set is one row in the Knowledge bases & documentation group:

On the rowWhat it tells you
NameOpens the editor for that knowledge base or documentation set.
Kind and countsKnowledge base or Documentation, then how many articles (or pages) it holds — drafts are counted too — and how many searchable pieces the AI has stored from it, for example 40 pieces indexed. One article usually becomes several pieces, and only published items are stored. Not indexed yet means nothing is stored.
LanguageA dropdown with the language the content is written in. It decides how the AI searches the text, and it translates nothing. You change it right on the row; a +n beside it means extra translations are configured.
On public site / AI onlyOn public site — visitors reach this one at your web address — or AI only. Only one knowledge base and one documentation set can be on the public site at a time.
⋯ menuEdit content, then either Visit the live page and Take down from the public site (on the one that is on the public site) or Show at followed by your address (on all the others), then Copy to another brand…, Move to another brand… and Delete.

When the brand has at least one knowledge base, an All articles button sits in the group's header. It opens a flat list of every article in this brand's knowledge bases.

Being on the public site is not the same thing as publishing.On public site and AI only decide which source visitors reach at your web address. Published is a separate switch on each individual article, on each documentation page and on each documentation category: it decides whether that item is shown at all, and whether the AI can read it. Knowledge base categories have no Published switch of their own. An unpublished article is a draft — it still exists, it is simply not shown and not indexed.

Publishing an article is what teaches the AI

Publishing indexes the article immediately. No chat widget and no AI agent are needed for it, so a brand with no chat bubble anywhere still builds up the knowledge its phone agent and its inbox suggestions answer from. Unpublishing removes it again, and editing a published article re-indexes just that article.

Who can open this screen

A team member needs any one of three rights: Content sources, Knowledge base or Documentation. Inside, each group checks its own — knowledge-base rows need Knowledge base, documentation rows need Documentation, and the other groups are shown only to Content sources holders. If your plan does not include the Documentation module, the Documentation card in the Add knowledge window is locked and says Available on a higher plan. instead of failing when you click it.

How much Yaplet can learn

Money decides what Yaplet may learn, and nothing already taught is ever removed.
  • On a paid plan there is no limit on how many pieces Yaplet learns (a website holds up to 2,000 pages).
  • During the free 14 days, or with bought credit and no plan, no limit shows on this screen, but one exists. If you reach it, the website row says You've reached your knowledge limit with a Write to us link.
  • With no paid plan, the last 500 credits are kept for answers: once fewer than 500 are left (the free 14 days' credits and bought credit together), Yaplet stops learning and keeps answering. A help article you publish then still goes live and is learned within the hour after you add credit; a documentation page still goes live too, but its new text is not learned, the payment window opens and Yaplet keeps the version it knew.
  • If nobody can pay (no plan, no bought credit, and the free 14 days are over), nothing is learned or refreshed and Yaplet's answers are paused. A help article you publish still goes live, and Yaplet learns it within the hour after someone pays.
  • Some older free workspaces keep a limit of 250 searchable pieces across the whole organization, shown on this screen as 212 of 250 on the free plan. Everything the AI reads draws on it: knowledge base articles, documentation pages, uploaded files, website pages and Q&A entries, in every brand. One article is usually several pieces. When it is full the screen says Your free plan's knowledge is full, and publishing new articles and adding new sources is paused until you remove something or upgrade.
See Free plan limits for the full breakdown.

Knowledge base or documentation

Yaplet gives you two kinds of structured content. They serve different readers, and a brand can hold both.

Knowledge base

Purpose: Customer-facing help articles in plain, conversational language

Content style: Text and sentences, written with flexible HTML editing

Editor: HTML Article Editor

Where people read it: Your public web address when it is the one being served, and the chat widget's Help tab

Best for: Support questions, FAQs, troubleshooting, product tutorials

Documentation

Purpose: Technical documentation with structured, developer-focused content

Content style: Formatted technical detail built from markdown components

Editor: Markdown Article Editor

Where people read it: Your public web address when it is the one being served

Best for: API references, technical specs, developer guides, structured manuals

Both feed the AI in exactly the same way, whether or not they are on your public address.

When to use a knowledge base

Choose a knowledge base when your readers are customers who want a quick answer to a common question.
  • Support questions: "How do I reset my password?" or "How do I upgrade my plan?"
  • Product tutorials: step-by-step guides to using your features
  • Troubleshooting: common problems and their fixes, in everyday language
  • FAQ-style content: anything you also want available inside a chat conversation

When to use documentation

Choose documentation when your readers are technical and need detailed, structured information about APIs, configuration or implementation.
  • API documentation: specifications, endpoints and code examples
  • Developer guides: integration instructions, SDK documentation, technical workflows
  • System configuration: detailed setup procedures
  • Technical specifications: component-based content that needs precise formatting

Key decision factors

  • Knowledge base: customers, end users, a non-technical audience
  • Documentation: developers, technical teams, system administrators :::
  • Knowledge base: flexible HTML articles, conversational tone, rich media
  • Documentation: structured markdown, technical components, code examples :::
  • Knowledge base: your public web address plus the chat widget's Help tab
  • Documentation: your public web address, as a technical portal :::
  • Knowledge base: fewer support tickets, self-service help
  • Documentation: developer onboarding, API integration, technical reference ::: ::
Many organizations keep both — a knowledge base for customers and documentation for developers — and link between them.

How each one is structured

Both are three levels deep, and the middle level is what a reader navigates by.

  • A knowledge base holds categories, and categories hold articles. Categories can sit inside one another.
  • A documentation set holds categories, and categories hold pages. A page can itself hold sub-pages, so one category can carry a whole nested tree.

Your plan does not ration the articles or pages you write. What your plan decides is whether the Documentation module is available to you at all. How many knowledge bases and documentation sets one brand can hold is a fixed limit per brand, explained on the Knowledge bases page.

Knowledge bases

Create a knowledge base, set its language, see how many pieces are indexed, and delete, copy or move it.

Categories and articles

Organize content into nested categories, publish articles, and read the rating feedback visitors leave.

Editing articles

Article settings, the AI writing tools, and what the Published switch actually does.

Documentation sets

The documentation side: create a set, organize its categories and pages, and control what is published.

Building docs

The split-pane markdown editor, the page tree, the ready-made components and the AI writing tools.

Share links

Hand any article, page, category or whole source to someone as a link — and which kind survives a rename.

Your public web address

What visitors find at your address, how you choose what is served there, and why taking something down deletes nothing.

Copying and moving content between brands

Copy makes an independent duplicate that arrives as drafts; Move hands the original over with everything it has learned.

::

Automate with AI: you can also create knowledge bases, write articles with AI-generated content and manage categories using Yaplet or any MCP-compatible client. Try prompts like "Create a knowledge base article about how to reset passwords" or "Set up a new knowledge base for our mobile app". See the Knowledge Base Tools reference for every available operation.

© 2026 Yaplet. All rights reserved.