खोजी · the seeker

Semantic site search that never touches a server.

A 2-line embed runs a tiny WASM embedding model in your visitor's browser — indexing your pages for meaning, not just keywords. Zero backend. Zero per-query cost. Which is why the price never moves, no matter how much traffic you get.

index.html — the entire integration
<script src="https://khoji.aiskillhub.info/widget/khoji.js"
  data-khoji async></script>

That's the whole install. Free for sites up to 200 pages.

↑ This is the actual widget, indexing the site you're on right now, in your browser. Try "how much does it cost" or "is my data safe".

✓ No account needed✓ No cookies, no tracking✓ Works on WordPress, Shopify & static sites

How a search engine fits in a script tag

01

Paste two lines

Drop the script tag before </body> on any site — WordPress, Shopify, Hugo, plain HTML. No build step, no API keys, no dashboard required to start.

02

Your visitor's browser does the work

Khoji reads your sitemap, indexes your pages, and runs a tiny quantized embedding model via WebAssembly — right on the visitor's device. Keyword search is instant; semantic kicks in seconds later.

03

Pay the same at 10 or 10 million queries

There is no search server. Every query is computed client-side, so our costs don't grow with your traffic — and neither does your bill. Flat price, forever.

Everything the expensive tools do.
Minus their invoice.

Semantic, not just keywords

“Refund my order” finds your “Returns policy” page. Meaning-based matching via 384-dim MiniLM embeddings, hybridized with BM25 for exact terms.

Nothing leaves the browser

Queries, page content, vectors — all stay on the visitor's device in IndexedDB. GDPR and DPDP compliance by architecture, not by paperwork.

14 KB embed

The widget is 14 KB gzipped-class. The 23 MB model loads lazily from a CDN only when someone actually searches, then caches.

Works offline after first index

Docs sites and PWAs get search that keeps working on flaky Indian mobile networks once the index is warm.

⌘K out of the box

A polished command-palette modal with keyboard navigation, dark mode, highlights, and your accent color. Shadow DOM — your CSS stays untouched.

No lock-in, no crawler to wait on

Content changes ship with your deploy. No re-crawl queue, no sync jobs, no webhook plumbing. Delete the script tag and you've fully uninstalled.

The honest comparison

KhojiAlgoliaTypesense Cloud
Setup2 lines, 2 minutesAPI keys + indexing pipelineCluster + sync pipeline
Cost at 100k queries/mo₹0 – ₹999 flatUsage-billed, grows with traffic~$22+/mo cluster, always on
Where queries runVisitor's browserTheir serversYour cluster
Data leaves your siteNeverYes — content + queriesYes — to your cluster
Semantic / meaning-basedYes (WASM embeddings)Yes (paid tiers)Yes (config required)
Best forSites & docs up to ~2k pagesLarge e-commerce catalogsTeams running infra

Honestly: if you have 50,000 SKUs and need merchandising rules, buy Algolia. If you have a content site, docs, or a store under 2,000 pages — you've been overpaying.

Flat pricing. The traffic spike is on us.

Free

₹0/mo · $0 global

For personal blogs and small sites.

  • 1 site
  • Up to 200 pages
  • Semantic + keyword search
  • ⌘K modal, dark mode
  • “Search by Khoji” badge
Most popular

Pro

₹999/mo · $12 global

For businesses and serious blogs.

  • 1 site
  • Up to 2,000 pages
  • Badge removed
  • Custom accent & placeholder
  • Priority email support

Studio

₹2,499/mo · $29 global

For agencies and multi-site owners.

  • 5 sites
  • 2,000 pages each
  • Search analytics (what visitors seek)
  • Badge removed
  • Priority support

Questions people actually ask

How is this different from Algolia or Typesense?

They run your index on their servers and charge per query volume or per record. Khoji computes everything in the visitor's browser, so we have no per-query cost to pass on — pricing stays flat. For sites under ~2,000 pages, quality is comparable and setup is 100x simpler.

Won't indexing in the browser be slow for visitors?

Keyword (BM25) search is available within a couple of seconds — pages are fetched in parallel from your own origin. The semantic model warms up in a background worker and upgrades results when ready. The index is cached in IndexedDB, so repeat visits are instant.

What does it need from my site?

A sitemap.xml (every WordPress, Shopify, Next.js, Hugo site has one). No sitemap? Khoji falls back to crawling same-origin links. That's it.

Does it work in Hindi and other Indian languages?

Keyword search works in any language today. Semantic search currently performs best in English; multilingual model support (Hindi, Tamil, Bengali) is on the paid roadmap — one attribute swap when it ships.

Is the free tier really free?

Yes — one site, up to 200 pages, with a small “Search by Khoji” badge. We can afford it precisely because your visitors' browsers do the compute, not our servers.

Your site could have this before chai gets cold.

Copy the snippet, paste it before </body>, reload. Done.

index.html — the entire integration
<script src="https://khoji.aiskillhub.info/widget/khoji.js"
  data-khoji async></script>

Need more than 200 pages? See Pro plans →