The Knowledge Base is the organizational context Enterpret holds about your product, so the Enterpret Agent answers in your own terms instead of generic ones. This article explains what lives in the Knowledge Base, how the agent reads from it during a session, how every answer is cited back to its source, and how to add knowledge in settings.
The Knowledge Base sits alongside the Context Graph. The Context Graph is your unified customer context: all your feedback plus the related objects around it, like accounts, deals, and users. It answers what is happening with your customers. The Knowledge Base supplies the product truth the agent needs to interpret that: what a feature is called, what shipped in the last release, how your team defines a tier or a plan. Together they let the agent ground an answer in your reality rather than its best guess.
What lives in the Knowledge Base
The Knowledge Base holds the context that explains your product and your organization, including, for example, uploaded documents and files such as a strategy memo or an onboarding deck, the support playbooks your team uses to resolve issues, PRDs that describe what a feature is meant to do, your product terminology and internal vocabulary, changelogs of what shipped and when, your help-center and API documentation, and the web links you submit for Enterpret to crawl and index.
How the agent uses the Knowledge Base in a session
Within a session, the agent queries the Knowledge Base automatically when a prompt depends on what something is named, what shipped, or how your team defines a term. It reads the relevant knowledge before it answers, and the step appears in the live work trail alongside its other steps.
A typical sequence:
You ask a question. "Summarize complaints about the new billing flow from the last 30 days."
The agent reads the Knowledge Base for what "the new billing flow" refers to in your product, and the changelog entry for when it shipped.
The agent queries the Context Graph for matching feedback in that window.
The agent returns a grounded answer that uses your feature name and frames the volume against the right release date.
Knowing your changelog, the agent separates a complaint about a recent change from a long-standing gripe; knowing your terminology, it resolves "the new billing flow" to the right feature instead of guessing from the words alone.
How answers are cited
Every answer is grounded in your data and cited back to its source, so you can check the evidence behind any claim. Those citations carry through into anything the agent generates, so a report it produces is as verifiable as the answer in the chat.
Adding and populating knowledge
Organization admins and editors manage the Knowledge Base from knowledge settings. Add the documents, playbooks, and docs that describe your product, and submit the links you want Enterpret to crawl. The richer the Knowledge Base, the more precisely the agent grounds its answers.
To add a document or file:
Open knowledge settings.
Add the document or file you want the agent to use.
Confirm it appears in the source list.
To submit a link for crawling:
In knowledge settings, add the URL you want Enterpret to fetch.
Enterpret crawls and indexes the page so the agent can read it.
Uploading files inside a session
Separately from the persistent Knowledge Base, you can attach flat files to a single session. The agent stages them into its sandbox and works over them with code, alongside the Context Graph. Supported formats are CSV, text, and data files, up to 10 MB per file and 20 live files per session. These attachments live with the session rather than persisting into the org's Knowledge Base.
Limitations
Limitation | Implication |
Knowledge is only as current as its sources | A changelog or doc that is out of date will ground answers in stale product truth. Keep sources current. |
Crawled pages reflect the last crawl | If a crawled page changes, the agent reads the indexed version until the next crawl. |
Coverage gaps surface as gaps | If the Knowledge Base lacks context for a question, the agent works from the feedback alone and may not resolve a term or release the way your team would. |
FAQs
What is the difference between the Knowledge Base and the Context Graph? The Context Graph is your unified customer context: feedback plus the related objects around it, like accounts, deals, and users. The Knowledge Base is the context that explains your product: docs, playbooks, PRDs, terminology, changelogs, and crawled links. The agent reads both in a session: the Context Graph for what is happening with your customers, the Knowledge Base for what your product is.
Do I have to tell the agent to use the Knowledge Base? No. The agent queries it automatically when a prompt depends on product context, and the step appears in the session work trail.
Can I submit a link instead of uploading a file? Yes. Add the URL in knowledge settings and Enterpret crawls and indexes the page for the agent to read.
Who can add to the Knowledge Base? Organization admins and editors. The Knowledge Base is shared across your whole organization, so what they add grounds the agent's answers for everyone.
Questions or feedback on the Knowledge Base? Reach out to the Enterpret team at and we will help.