Build a knowledge base chatbot and AI help center assistant from files, sites, Q&A, and 27+ connectors—retrieval augmented generation for customer support so the AI chatbot uses your documentation, not generic guesses.
In the app, open Knowledge Base from the dashboard sidebar.
Add content
Use the upload workspace for files and sites, add structured Q&A, and turn on connectors so help docs, CRM notes, files, and calendars stay fresh without copy-paste.
Drop in the materials customers and staff already have: documents, images, audio, video, and archives. Content is tracked as sources you can review and replace.
Point at a help site, docs portal, or marketing domain and crawl with sensible limits—so public pages land in your knowledge base under your control.
Capture exact questions and approved answers—ideal for policies, pricing snippets, and anything you want repeated verbatim.
Connect Notion, Zendesk, Confluence, Google Drive, Shopify, Salesforce, calendars, and many more. Scheduled jobs pull updates into the knowledge bases you choose.
Structure
Create one or many knowledge bases, mark whether they are general or tied to a specific agent, and assign only the right scope so assistants do not overreach.
Split internal runbooks from customer-facing help, or separate brands and regions—each base has its own sources and documents.
Each assistant only searches the knowledge bases you attach. That keeps HR answers out of a storefront bot and vice versa.
Every document ties back to a source—upload, crawl, connector, or Q&A—so you can trace where an answer came from and refresh or remove it confidently.
Retrieval & replies
Questions are matched to the best passages with similarity search, with fallbacks when matches are thin. The model then answers using that context and your agent rules.
Long material is split into searchable pieces, embedded for meaning-based lookup, and stored with your organization and knowledge-base boundaries enforced.
Retrieval can weigh recency, quality hints, and the kinds of documents you want promoted—so the right passages surface first.
The assistant model receives retrieved context plus your system instructions. Replies can stream for a snappy widget or return in one block for integrations.
In the product
The Knowledge Base area in the dashboard is built around documents, ingestion, search, and visibility into how content is used.
Browse and manage indexed material, see status, and jump to the knowledge base or source behind each item.
Central place to add files, start site crawls, and monitor progress—aligned with the same actions the APIs expose.
Look across what is indexed before customers do—sanity-check coverage and find gaps in your content.
Where enabled for your plan, see how knowledge is used and inspect embedding-oriented views to understand indexing health.
Integrations
Names below come straight from the integration manifest—what you read here matches connector pickers and marketing elsewhere. Each family syncs into knowledge bases so agents stay aligned with the apps your team already trusts.
9 highlighted integrations in this group—sync content into knowledge bases your agents can search.
4 highlighted integrations in this group—sync content into knowledge bases your agents can search.
4 highlighted integrations in this group—sync content into knowledge bases your agents can search.
5 highlighted integrations in this group—sync content into knowledge bases your agents can search.
5 highlighted integrations in this group—sync content into knowledge bases your agents can search.
27 named integration paths across documentation, commerce, CRM, files, and calendars—each can feed knowledge bases your agents search.
See how knowledge fits the rest of the stack on the features overview, compare limits on pricing, or talk to us on contact.
Next step
Start a workspace, create a knowledge base, add uploads or a connector, and wire the base to the assistant your customers will meet first.