> ## Documentation Index > Fetch the complete documentation index at: https://docs.helpalive.com/llms.txt > Use this file to discover all available pages before exploring further. # Train the Agent on your docs and content > Train the HelpAlive Agent on your documentation. Crawl a URL, upload PDFs and Word documents, or paste your own custom answers. Setup takes minutes. The **Agent → Train** tab is where you give your Agent the knowledge it answers from. Point HelpAlive at your existing content — a docs site, files, or pasted text — and we handle the rest. ## What you can add | Source | How to add | Best for | | ------------------------- | ---------------------- | -------------------------------------------------- | | **Docs site** | Crawl from a URL | Existing public help center or product docs | | **PDF** | Upload a file | Onboarding guides, internal manuals, contracts | | **Word documents (DOCX)** | Upload a file | Knowledge articles your team writes in Word | | **Markdown** | Upload or paste | Technical docs you keep alongside your code | | **Plain text** | Paste in the dashboard | Short FAQs, custom answers, special-case responses | ## Adding a source Open your dashboard and go to **Agent → Train**. From there: ### Crawling a URL Use this when you already have a docs site or help center. Configure: * **Start URL** — the page to begin from. * **Crawl depth** — how many link-hops deep to follow. * **Stay on domain** — when on, the crawler ignores cross-domain links. * **Include / exclude paths** — narrow the crawl with patterns like `/blog/**` or `/api/v2/*`. Crawls run in the background. Most sites finish in a few minutes; you can keep working while they run. ### Uploading files Drag and drop one or many PDFs, DOCX files, or Markdown files. Each file becomes its own source with metadata you can manage. Re-uploading the same file replaces the existing version — no duplication. ### Pasted text For short content — FAQs, plan-specific answers, internal policies — paste it directly. Useful when something isn't in your public docs but you still want the Agent to know about it. ## Managing sources Each source shows: * **Status** — `crawling`, `indexing`, `ready`, or `error`. * **Pages or sections** — what was discovered. * **Last indexed** — useful for re-crawling stale content. * **Publish status** — `draft` or `published`. You can re-crawl, archive, or delete a source any time. Deleting removes its contribution to Agent answers immediately. ## How the Agent finds answers You don't have to think about how the Agent searches your content — but here's the short version. HelpAlive reads everything you add and stores it so the Agent can find the right passage in milliseconds. Whether your user phrases the question like your docs or in their own words, the Agent finds the closest match and grounds its answer in it. Just keep your sources up to date and the Agent will keep up. ## Draft and publish Knowledge sources go through the same draft / publish workflow as everything else in the Agent. New or changed sources stay in **draft** until you publish — your users see no change until then, but you can preview them on **Agent → Test**. The Test tab intentionally uses your **draft** content so you can verify a new source improves answers before shipping. See [Publishing](/agent/publishing) for the full workflow. ## Quality monitoring The **Studio → Topics** inbox flags every question the Agent couldn't answer well — usually because something is missing from the knowledge base. Topics are grouped, ranked, and surfaced so you know what to write next. See [Topics](/studio/topics). ## What's never indexed * **Private content behind login** — the crawler only follows publicly accessible URLs unless you upload private content directly as files. * **Drafts you haven't published** — draft sources never appear in production answers. * **Customer data from your end users** — events and conversations are stored separately from knowledge. ## Next steps Match the Agent to your brand. Stage and preview before going live. Find what to write next.