Data sources: train your chatbot on your website, PDFs and docs

Train your Helpnest agent on website URLs, PDFs, Markdown, CSV, HTML, JSON, pasted text, and Q&A pairs. Learn about limits, re-syncing, and training status.

Your agent only answers from the sources you give it. Add sources from the Data sources tab of an agent.

Source types

TypeBest forDetails
WebsiteHelp centers, marketing sites, docsEnter a URL. Helpnest follows links on the same site and reads up to 25 pages by default (you can set 1 to 200).
FilePolicies, manuals, price listsUpload one or more files. See the supported types below.
TextContent that isn't onlinePaste up to 500,000 characters with a title.
Q&AExact answers to specific questionsWrite a question and the answer you want given. Use this to correct or pin important answers.

Supported file types

PDF, TXT, Markdown (.md, .markdown), CSV, HTML, and JSON. Scanned PDFs without selectable text can't be read. The maximum file size depends on your plan: 5 MB on Free, 10 MB on Startups, 25 MB on Growth.

Training status

New sources are processed in the background. While a source is training, the agent may not know its content yet. If training is interrupted, the source shows an error and you can retry it.

Re-sync and delete

When your website changes, use Re-sync on the source to crawl it again. Deleting a source removes its content from your agent's answers immediately.

Knowledge limits

Knowledge is measured in training pages of 2,000 characters, shared across the agents in your workspace.

PlanTraining pagesFile uploads per monthMax file size
Free100105 MB
Startups50010010 MB
Growth2,50040025 MB

Tips for better answers

  • Start with your help center or FAQ page. They match how customers ask questions.
  • Add Q&A pairs for answers that must be exact, like refund windows or prices.
  • Remove outdated sources rather than adding corrections on top of them.
  • Use custom instructions for rules, not facts. Facts belong in sources.

Last updated October 30, 2026