Connect your knowledge, keep it in sync
Add a URL, Confluence space, OneDrive, or Notion workspace. Kelu discovers every page, extracts clean content, and automatically re-indexes when content changes. No code, no webhooks, no manual uploads.
How the knowledge crawler works
Built for the real-world complexity of enterprise knowledge sources
Sitemap Discovery
Kelu fetches sitemap.xml first, then follows links to discover every page the crawler should index. No manual URL lists required.
JavaScript Rendering
Turn on Render JavaScript to load pages in a headless browser before extraction. SPAs, Next.js sites, and dynamically rendered intranet pages all index accurately.
Clean Markdown Extraction
Strips nav, footers, ads, and boilerplate. Preserves code blocks, tables, and heading hierarchy — the signal that matters for RAG retrieval.
Incremental Sync
Every source type re-syncs on its own cadence, from 5 minutes to 24 hours, and GitHub repos can reindex on every push. Only pages that changed are re-embedded, keeping costs and latency low.
Your index stays current automatically
Stale knowledge is worse than no knowledge — teams get wrong answers. Kelu polls each source on its own fixed schedule and re-indexes only the pages that changed, keeping every answer grounded in the latest content.
24 source types, in the categories the product uses
This is the whole catalogue — the same list, in the same order, that you see when you click Add source. Each card links to its setup guide.
Documentation & content
7Crawl a documentation site, sitemap and robots.txt aware
Confluence Cloud spaces
Workspace pages and databases
READMEs, docs/ folders and changelogs, by convention
Source code, chosen by language, directory and regex
Video transcripts from a channel and optional playlists
Answers you write by hand when there is no page to crawl
Community & conversations
3
Issues & tickets
4
Customer support
5
Storage & files
4
Developer & API
1
Five of these read one GitHub repository and are deliberately separate: wanting the issue archive is not wanting every pull request review thread with it. There is also a way in that is not a source type — the browser extension saves a page, or a passage you highlighted, straight into the index. Slack and Discord can no longer be added as sources; ones added earlier keep syncing.
Safe syncs that keep your knowledge base intact
A failed sync can't empty your knowledge base
If one sync would remove more than 45% of a source's pages, for example during a docs outage, Kelu holds the removals until someone on your team approves them.
Unchanged pages are skipped
ETag checks and content hashes mean a page that hasn't changed is never embedded again.
Encrypted credentials
OAuth tokens and connector secrets are sealed with AES-256-GCM. If one can't be opened, the sync stops instead of running without it.
Each source on its own schedule
Tickets and Notion every 5 minutes, Confluence and Google Drive every 10, GitHub repos hourly, websites daily.
Got questions?
Frequently Asked Questions
Common questions about Kelu knowledge source connectors
Yes. Kelu respects robots.txt directives by default. You can also configure a custom crawl rate and explicit allow/deny URL patterns in the source's settings.
Not with a login. The crawler has no sign-in settings; for sites that allow only known IPs, set an Outbound proxy. For content behind a login, upload it as files instead.
Turn on Render JavaScript and Kelu renders each page before extracting content. SPAs, Next.js sites, and any other JS-rendered platform are supported.
Most knowledge sources of 100–500 pages complete in under 2 minutes. Larger sources (1,000+ pages) typically finish within 10–15 minutes.
Yes. Configure URL patterns to exclude (e.g., /changelog/*, /blog/*) in the source's settings. You can also provide an explicit allow-list of URL prefixes to crawl.
Index your first knowledge source in minutes
Connect a source, click sync, and your knowledge base is ready.
Free to start · No credit card required