Picking site search in 2026 is harder than “just use Algolia” suggests

I compared four products that all power “site search” but solve different problems: Pagefind, Algolia, Typesense Cloud, and Orama. The forum answers mostly skip the part that actually matters: where search runs, what you get billed for, and whether the index can stay private.

For a public static site (docs, blog, marketing pages, open knowledge base), my default is Pagefind. It indexes your built HTML after deploy, ships a chunked static bundle to the CDN, and runs search in the browser. No SaaS account, no query metering, no search server. Pagefind’s own docs claim a 10,000-page site can search with under 300 kB of network payload, often closer to 100 kB. The tradeoff is architectural: the index goes to the client, so this is wrong for content that must stay hidden.

When I want a managed service with analytics, query suggestions, merchandising controls, and the broadest commercial search tooling, I still look at Algolia first. The Build plan is free to test with 10,000 search requests/month and 1 million records. Production pricing on Grow ties overages to requests ($0.50 per additional 1,000 on Grow) and records ($0.40 per additional 1,000 after 100,000). Algolia says most queries process in 1-20 ms across 70+ data centers in 17 regions. The catch is pure SaaS economics: no self-hosting, and costs climb with traffic and index size.

Typesense Cloud is the managed option I reach for when query volume is high and I want billing that looks more like infrastructure than usage counters. Dedicated isolated clusters, region selection, optional HA and Search Delivery Network, and no per-search or per-record charges. You pay hourly cluster fees plus bandwidth (usually about $0.09-0.12/GB). The one-time free tier gives 720 hours and 10 GB bandwidth on select configs. The open-source Typesense engine is the exit hatch if you outgrow the cloud SKU.

Orama is the wildcard for TypeScript-heavy teams. The Apache 2.0 library runs in browser, server, or edge with full-text, vector, and hybrid search plus a plugin ecosystem. Orama Cloud and hosted search exist, but public pricing and security detail for the search product are much thinner than Algolia or Typesense. I would not buy Orama Cloud on a procurement checklist from public docs alone. I would absolutely try the OSS library in an Astro or VitePress docs site.

You cannot compare these on features alone

Every product here can put a search box on your site. The useful split is what happens when someone types.

The questions I ask before picking one:

You will not get a perfect apples-to-apples table. Use the numbers below as directional, then test the path you actually need: faceted browse, typo tolerance, vector retrieval, or click analytics.

How the architectures differ

flowchart TD
    A[Site content] --> B{Search architecture}

    B --> C[Pagefind]
    C --> C1[Run Pagefind after build]
    C1 --> C2[Static index bundle on CDN/host]
    C2 --> C3[Search runs in browser]

    B --> D[Algolia]
    D --> D1[Push/crawl content into Algolia indices]
    D1 --> D2[Algolia SaaS clusters + analytics]
    D2 --> D3[Frontend queries Algolia APIs]

    B --> E[Orama]
    E --> E1[Orama JS in browser/server/edge]
    E --> E2[Orama Cloud managed project]
    E1 --> E3[Local/self-managed execution]
    E2 --> E4[Hosted search endpoint]

    B --> F[Typesense Cloud]
    F --> F1[Sync data via REST/ETL]
    F1 --> F2[Dedicated cloud cluster]
    F2 --> F3[Optional HA + Search Delivery Network]
    F3 --> F4[Frontend/backend queries Typesense]

Privacy follows architecture. Pagefind has no third-party query processor by default, which is great for public docs and terrible for confidential content because users download the index. SaaS engines can filter results server-side, scope API keys, and enforce permissions at query time. That is the pattern you need when search spans authenticated or personalized data.

Free tiers at a glance

Two things to keep in mind. Algolia’s Build plan overages are less clearly exposed on pricing pages than Grow’s. Orama.com has a broader AI platform pricing page ($20/month, 20M tokens) that does not map cleanly to site search; I am not treating that as a site-search matrix without clearer official linkage.

What the free tiers actually feel like in use

Pagefind: delete the search backend

Pagefind is the option I would ship on any public static site before I reach for a SaaS engine. Run it after your SSG build, drop the output next to your HTML, wire up the UI component or custom JS hook, done. It works with any static HTML output, supports filtering, sorting, highlighting, multilingual indexes, custom records via Node/Python APIs, and faceted browse mode.

The performance story is about payload, not millisecond SaaS latency claims. Chunked indexes mean users download only what they need. Real-world demos include MDN, Godot, and XKCD. For a docs site that already rebuilds on every content change, adding Pagefind to the pipeline is usually one command and a script tag.

What you do not get: hosted analytics dashboards, managed synonyms/merchandising comparable to Algolia, personalization, or protected query-time filtering. Every content update means rebuild and redeploy the index. If that cadence is fine and the content is public, Pagefind removes an entire vendor category from your stack.

Algolia: the turnkey commercial default

Algolia is still the product I recommend when search is revenue-critical and the team wants polish over cost minimization. Replicas for alternate ranking, custom ranking, synonyms, typo tolerance, faceting, analytics, click/conversion events, query suggestions, and the InstantSearch widget ecosystem cover most commercial site-search needs out of the box. SDK coverage is unusually broad.

The structural tradeoffs are also clear. No public on-prem or FedRAMP service. Production pricing follows requests and records, which gets expensive or at least harder to forecast at scale. Algolia documents temporary operational backups but explicitly tells customers to build their own backup solutions. For ecommerce, multi-region catalogs, DocSearch-style documentation with autosuggest, and teams that want one vendor for search plus merchandising optimization, Algolia is the safest enterprise SaaS answer in this set. It is not the cheapest default.

Typesense Cloud: managed with an open-source exit

Typesense Cloud is the managed engine I keep shortlisting when Algolia’s metering model makes me nervous. Dedicated isolated clusters, region pinning, full-disk encryption, HTTPS-only, optional HA and Search Delivery Network for geo-distributed latency. Feature-wise it now covers keyword search, synonyms, typo tolerance, faceting, analytics, vector/hybrid/semantic search, and conversational retrieval patterns.

The InstantSearch adapter and Algolia migration docs make frontend reuse practical. Customer testimonials on the cloud site cite lower costs after migration, quick InstantSearch reuse, and deployments at Kick.com, MacroFactor, and others. I would not call it frictionless for beginners: you still think in cluster RAM, vCPU, and region topology. Pricing lives behind an interactive calculator rather than a simple “starts at” table. Fewer enterprise-marketing bells than Algolia’s broader platform, but the core search feature set is strong and the open-source engine is a real strategic hedge.

Orama: composable TypeScript search, murky hosted story

Orama is the most interesting pick if your stack is already TypeScript end to end and you want search as a library, not just a hosted search box. Full-text, vector, hybrid, typo tolerance, facets, filters, geosearch, boosting, grouping, sorting, result pinning, an answer engine, and plugins for analytics, secure proxying, embeddings, persistence, and docs frameworks (Astro, Docusaurus, Nextra, VitePress). Runtime options span browser, server, edge, and managed cloud.

The problem is not capability. It is public clarity. Hosted search, Orama Cloud, and the broader Orama.com AI platform do not present a comparably clean pricing and compliance page to what Algolia and Typesense publish. I treat managed-service pricing and some enterprise controls as unspecified from public search docs rather than guessing. For experimentation, modern docs-site integration, or hybrid/vector retrieval baked into TS code, Orama is compelling. For conservative procurement today, Typesense Cloud or Algolia are easier to evaluate and buy.

What I would pick

flowchart TD
    start[Need site search] --> publicQ{Public static HTML?}
    publicQ -->|Yes| analyticsQ{Need hosted analytics or protected filtering?}
    analyticsQ -->|No| pagefind[Pagefind]
    analyticsQ -->|Yes| managedA[Algolia or Typesense Cloud]
    publicQ -->|No| managedQ{Want fully managed?}
    managedQ -->|Yes| dxQ{Polished SaaS DX top priority?}
    dxQ -->|Yes| algolia[Algolia]
    dxQ -->|No| typesense[Typesense Cloud]
    managedQ -->|No| tsQ{TypeScript-heavy stack?}
    tsQ -->|Yes| orama[Orama]
    tsQ -->|No| other[Consider self-hosted Typesense or other engines]

Public docs or blog on a static site generator: Pagefind. Add it to the build pipeline before you sign up for anything with a dashboard.

Ecommerce or revenue-critical discovery with analytics and merchandising: Algolia, if the request/record bill is acceptable. Typesense Cloud if query volume is high and you want infra-style economics with an open-source fallback.

SaaS product with auth-gated or personalized search: Typesense Cloud. Server-side filtering, dedicated clusters, and clear security documentation beat shipping an index to the browser.

Astro/VitePress/Docusaurus site where you want vector or hybrid search in TS: Orama OSS first. Evaluate Orama Cloud only after you have confirmed pricing and compliance with their team.

Migrating off Algolia but keeping InstantSearch UI patterns: Typesense Cloud with the InstantSearch adapter.

For my own static properties I would default to Pagefind, keep Typesense Cloud on the shortlist for anything dynamic or protected, and treat Algolia as the “we have budget and want the most turnkey commercial stack” option.

Gaps in the public docs

Orama Cloud shows up in the docs, but hosted search pricing and search-specific compliance detail were missing enough that I could not build a matrix comparable to Algolia or Typesense.

Typesense Cloud cluster prices depend on RAM, vCPU, and region. The billing model and free tier are clear. A single canonical “starts at” dollar amount is not.

Throughput numbers here mostly come from vendor docs, testimonials, and examples. I did not find an independent benchmark that ranks all four on equal footing.

Algolia Build plan overages are less explicit on public pricing pages than Grow’s tiers. Check inside the product before you extrapolate from Grow numbers to a Build deployment.

Pagefind is an architectural alternative, not a drop-in hosted SaaS search replacement. Great for public static content. Wrong when the index itself must stay confidential.

Sources

Pagefind

Pagefind homepage
Pagefind docs
Pagefind GitHub

Algolia

Algolia pricing
Algolia documentation
Algolia security and compliance
Algolia performance

Typesense Cloud

Typesense Cloud
Typesense documentation
Typesense Cloud security
Typesense Cloud pricing

Orama

Orama documentation
Orama GitHub
Orama
Orama Cloud