Algolia alternatives: which search engine to pick and how to move
Six engines broken down, one per reason to leave, and an eight-step migration plan.
The short answer: Typesense if you want to keep your InstantSearch frontend and stop paying per request, Meilisearch for the simplest setup and a flat bill, Elasticsearch if search is a serious engineering project, Doofinder for a store with no developer, AddSearch for a content site you want crawled, and Pagefind if your site is static and you're done paying for search. One thing before you pick: Algolia doesn't let you export records from the dashboard, only through the API or CLI. We compared these six destinations and their migration paths in 2026.
The ranking is editorial: a sponsored spot is always labeled on its card, and no affiliate link changes a rating.
Why the Algolia bill catches people off guard
Nobody leaves Algolia because search got bad. It's still fast, and its instant search is among the best out there. Teams leave because of how it counts, and because of what sits behind the higher tiers. Six reasons, checked against the 2026 price grid.
Every keystroke is a request
Algolia bills search requests, and search-as-you-type sends one per keystroke. A user on the Hugo forum put it plainly: one real search for "hugo algolia" counts as about ten. Past the 10,000 requests included each month, Grow charges $0.50 per 1,000.
Sort orders multiply your records
Sorting by price or by date means replicas, and every pre-sorted replica is a full copy of the index that counts toward your records. A 40,000-product catalog with three sort replicas weighs 160,000 records. Past the 100,000 included, each extra 1,000 costs $0.40.
AI is a separate price tier
AI Ranking, AI Synonyms and advanced personalization live on Grow Plus, where requests jump from $0.50 to $1.75 per 1,000. NeuralSearch, the semantic layer, is Elevate only, on an annual custom contract. Turning on the smart features multiplies the request price by 3.5.
Ten rules per index
Merchandising rules (pin a product, hide an out-of-stock item, redirect a query) cap at 10 per index on Free and Grow. You need Grow Plus to get 10,000. A store that runs promos through search hits ten fast.
A bill nobody can forecast
It's the top complaint in no-code communities: a FlutterFlow builder worried that 50 active users running ten searches each could spiral, and nobody could say by how much. On the Bubble forum, one team said it replaced roughly £1,000 a month of Algolia with a £20 Typesense server, at the cost of wiring the API calls itself.
No self-hosted option
Algolia is SaaS only: you can't run it on your own servers or in your own VPC. Free, Grow and Grow Plus run in the US, UK or EU West, and analytics retention stops at 7 days on Free and 30 on Grow. If you want your data and search logs in-house, you have to move.
You're leaving because… go to…
The right Algolia alternative depends on what's pushing you out. Usually one line fits. Each card jumps to the full breakdown below.
Its adapter reads Algolia-style filters.
Your bill swings every monthMeilisearchCloud from $20 a month, or free to self-host.
Search is a heavy engineering projectElasticsearchEvery knob exposed, if you have the team.
You run a store with no developerDoofinderShopify, BigCommerce or WooCommerce plugin, from $49 a month.
You want a content site crawled for youAddSearchA crawler indexes your pages, nothing to push.
Your site is static and you want to pay nothingPagefindIndex built at deploy, search runs in the browser.
Algolia alternatives, engine by engine
For each engine: why move there, who it's for, its catch, how to migrate from Algolia, and what it costs. Prices are what each vendor lists in 2026.
The engines
Typesense: keep InstantSearch, pay for the server
Shortest frontend migrationWhy move to Typesense. It works the most like Algolia and has the best migration tooling. Its InstantSearch adapter accepts facetFilters and numericFilters in Algolia format and reads _geoloc as is. Typesense Cloud doesn't bill per request or per record: you pay for a dedicated cluster by the hour, based on the RAM you pick, plus bandwidth at roughly 9 to 12 cents per GB.
Who it's for. Teams with an InstantSearch frontend (JS, React, Vue or Angular) they want to keep. No-code and mobile apps with spiky search traffic.
Who it's not for. A huge catalog on a tiny budget. Typesense holds the whole index in RAM, so memory is your data ceiling.
The catch. The cluster bills every hour it runs, traffic or not. Weigh that fixed cost against your current Algolia bill before committing. Self-hosted (GPL-3.0), pin a specific version in production instead of latest.
How to migrate. Export records with the Algolia CLI, define a schema (Typesense wants fields declared up front), import as JSONL. On the frontend, swap the search client for the adapter's. Two rewrites: filters becomes filter_by syntax, and sorting uses an index name like products/sort/price:asc.
- Free
- self-hosted (GPL-3.0)
- Paid
- Cloud: hourly cluster + bandwidth at $0.09 to $0.12 per GB
- Best for
- existing InstantSearch frontend
- Catch
- index lives in RAM
Meilisearch: the bill you can actually read
Easiest setupWhy move to Meilisearch. You keep what Algolia did well (typo tolerance, instant results, facets) with pricing you can predict. Cloud starts at $20 a month, with two ways to pay: usage-based (their example: $30 a month for 100,000 documents and 50,000 searches) or resource-based (a 0.5 vCPU, 1 GB instance at $18 plus disk). No per-seat fees.
Who it's for. SaaS products, marketplaces and content sites that want fast search without watching a request meter, and teams that want MIT-licensed software they can self-host.
Who it's not for. A big store that lives on merchandising: there's no direct equivalent to Algolia Rules, and the official migration guide skips them.
The catch. Settings and search parameters are strictly separate: an index setting can never be passed per query, and vice versa. A few Algolia parameters have no match at all (aroundPrecision, for one). Review each setting instead of pasting your JSON.
How to migrate. Meilisearch publishes an Algolia migration guide with JavaScript, Python and Ruby examples. InstantSearch doesn't plug in directly; you add the instant-meilisearch package. Good news for storage: no replicas, you declare sortableAttributes and sort at query time.
- Free
- self-hosted (MIT), 14-day Cloud trial
- Paid
- Cloud from $20/month
- Best for
- SaaS, content, marketplaces
- Catch
- no Rules equivalent
Elasticsearch: every knob, nothing prebuilt
Most powerfulWhy move to Elasticsearch. If Algolia's lock-in is the problem (no self-hosting, limited control over ranking), Elasticsearch is the opposite: analyzers, relevance scoring and vector search are all yours to tune. On Elastic Cloud Serverless, search starts at $0.09 per VCU-hour and storage at $0.047 per GB a month. Elastic's own example puts a 2 GB dev project at $24 a month and a 20 GB production one at $190 (On-demand pricing).
Who it's for. Engineering teams already running Elastic for logs or observability, and products where search is the product.
Who it's not for. A site that just needs a fast search bar. It's usually overkill.
The catch. Nothing comes ready: typo tolerance, autocomplete and facets are built query by query. VCU-hour billing is as hard to forecast as Algolia's until you have a month of history.
How to migrate. There's no Algolia importer: export NDJSON, write a mapping, load it through the bulk API. If you already automate with n8n, the n8n Elasticsearch node lets you push documents without a script.
- Free
- Elastic Cloud trial
- Paid
- from $0.09 per VCU-hour (dev example: $24/month)
- Best for
- engineering teams, custom relevance
- Catch
- everything is DIY
Doofinder: store search without a developer
Plug-and-play ecommerceWhy move to Doofinder. Algolia hands you building blocks. Doofinder hands you finished store search. It installs as a plugin on Shopify, BigCommerce, WooCommerce, Magento or PrestaShop, indexes your catalog on its own, and adds recommendations on top. Basic is $49 a month ($44 billed yearly) for 10,000 requests, with a 15-day trial and no card.
Who it's for. Shopify and BigCommerce merchants who had an agency set up Algolia and can't evolve it themselves.
Who it's not for. A SaaS app or a docs site. It's built for product catalogs.
The catch. Tiers climb fast: 10,000 requests on Basic, 150,000 on Pro ($149 a month), 400,000 on Advanced ($349). Category merchandising is Advanced only.
How to migrate. No product export needed: Doofinder reads your catalog from the store. The real job is pulling Algolia out of the theme. If you're changing platforms too, our guides to switching from Shopify and switching from WooCommerce cover the catalog move. One Shopify merchant reported uninstalling a search app whose leftover theme code kept breaking search for customers, so check your theme files after the uninstall.
- Free
- 15-day trial
- Paid
- Basic $49/month ($44 yearly)
- Best for
- store with no developer
- Catch
- 10,000 requests on Basic
AddSearch: a crawler does the indexing
Content sitesWhy move to AddSearch. With Algolia, every piece of content goes in through the API or a configured crawler. AddSearch starts from the crawler: point it at your domain, it indexes your pages, you drop in a search box. Keyword Search starts at $119 a month billed yearly, with a 14-day trial.
Who it's for. Content sites (media, universities, public sector, corporate sites) run by marketing teams without a full-time developer.
Who it's not for. Apps or stores that need fine-grained facets on structured data.
The catch. It's the priciest entry point of the six, it's an annual commitment, and AI Answers starts at $700 a month.
How to migrate. Nothing to import: register your domain, let the crawler run, swap the Algolia widget for AddSearch's. Recreate synonyms by hand from your Algolia export.
- Free
- 14-day trial
- Paid
- from $119/month billed yearly
- Best for
- content sites, no developer
- Catch
- annual commitment
Pagefind: no server, no bill, for static sites
FreeWhy move to Pagefind. Plenty of blogs and docs sites run Algolia for one search bar. Pagefind builds the index when you deploy and runs search in the visitor's browser. No server, no account, no metered requests. Free and open source.
Who it's for. Sites built with any static site generator (Hugo, Astro…): blogs, docs, marketing sites.
Who it's not for. Anything whose content changes without a rebuild, like a database-backed catalog. The index is only as fresh as your last deploy.
The catch. No search analytics and no merchandising rules. You won't know what people type unless you add tracking yourself.
How to migrate. Remove the Algolia script, add Pagefind to the end of your build, drop in its UI. Algolia DocSearch is still free for open-source project documentation and technical blogs, on application, with the "Search by Algolia" logo required, so keep it if you qualify.
- Free
- yes, open source
- Paid
- no paid plan
- Best for
- blogs, docs, static sites
- Catch
- index updates at build only
The 6 alternatives side by side
Our default picks: Typesense if your InstantSearch frontend has to survive the move, Meilisearch if you want a flat, readable bill. For scale, take 200,000 search requests a month on Algolia Grow: the 190,000 past the free 10,000 cost $95, or $332.50 on Grow Plus. One row per engine below: what's free on the left, entry price and migration path on the right.
| Free | Entry price and migration | |
|---|---|---|
| AlgoliaWhere you start | Free: 10,000 requests and 50,000 records a month, 10 rules | Grow: $0.50 per 1,000 requests past 10,000; Grow Plus: $1.75 |
| TypesenseInstantSearch frontend | Self-hosted (GPL-3.0) | Cloud: hourly cluster + bandwidth at $0.09 to $0.12 per GB; InstantSearch adapter |
| MeilisearchEasiest | Self-hosted (MIT), 14-day Cloud trial | Cloud from $20 a month; official Algolia guide, instant-meilisearch |
| ElasticsearchMost powerful | Elastic Cloud trial | From $0.09 per VCU-hour (dev example $24 a month); bulk API import |
| DoofinderEcommerce | 15-day trial, no card | Basic $49 a month (10,000 requests); store plugin, nothing to export |
| AddSearchContent sites | 14-day trial | From $119 a month billed yearly; crawler indexing |
| PagefindStatic sites | Free, open source | No paid plan; index added at build |
Entry prices before tax, as listed by each vendor in 2026, subject to change. Always run the numbers on your real request volume.
Moving off Algolia in eight steps
Order matters. Each step protects something: your data, your result ranking, the revenue that flows through search.
The steps
Map everything that touches Algolia
Day 1List every index, its sort replicas, facets, rules and synonyms, and everything that queries Algolia: site search, autocomplete, the mobile app, a Shopify connector, an automation. Also list what writes to the index: a script, a plugin, a Make scenario connected to Algolia.
Then export your top searches and no-results queries from analytics. They vanish with the account, and they're the best tuning data you have.
- Time
- half a day
- Risk if rushed
- a forgotten writer keeps feeding the old index
Export records, settings, rules and synonyms
Day 1The dashboard won't export records. Use the CLI: algolia objects browse dumps records as NDJSON, and algolia indices config export dumps settings, rules and synonyms into one JSON file. Repeat per index, replicas included if their settings differ.
Store the files outside Algolia. No active API key, no export.
- Tool
- Algolia CLI
- Format
- NDJSON and JSON
Translate settings into the new engine
Day 2No engine reads Algolia's JSON as is. Map setting by setting: faceting attributes become filterable attributes in Meilisearch and facet fields in a Typesense schema. Typo tolerance, stop words and synonyms almost always have a match; some geo settings don't.
Keep a two-column sheet: Algolia setting, new equivalent or "dropped". It's your proof nothing went missing by accident.
- Trap
- passing an index setting per query
Replace sort replicas
SavingsIn Algolia, every pre-computed sort was a copy of the index. Meilisearch sorts at query time on declared fields; Typesense takes the sort in the index name. Drop the replicas and your stored volume shrinks back to your real catalog size.
- Gain
- one less index copy per sort
Swap the frontend client, keep the widgets
Days 2 to 3If you use InstantSearch, don't rewrite it. Typesense has its adapter, Meilisearch has instant-meilisearch. You swap the search client and keep your components. Check each widget you use against the adapter's supported list.
Keep a 200 to 300 ms debounce. Even without per-request billing, your server will thank you.
- Risk
- widget not supported by the adapter
Rebuild merchandising rules by hand
ManualMerchandising rules don't migrate. Rebuild them one by one from your export: a pinned product becomes a boost field or sort, a redirect becomes a site-side rule. Use the chance to delete dead ones; on an old account, many point at promos that ended long ago.
- Time
- depends on rule count
Run both engines side by side
1 to 2 weeksWrite to both indexes for one to two weeks and compare results on your top 50 queries and your no-results list. Fix gaps before cutover, not after.
- Check
- top 50 queries
Cut over, watch, then close Algolia
Last daySwitch the frontend, then stop writes to Algolia. Watch the no-results rate and click-through for a week. Only then revoke API keys and close the account. If Algolia came in through a store app, confirm its code is gone from the theme.
- Watch
- no-results rate
Is your tool missing from this page? Email us: hello@hackceleration.com
What makes an Algolia exit go wrong
The same mistakes show up project after project, in forums and in vendor guides alike. None is hard to avoid. They all come from something nobody wrote down. The other guides in our switching series follow the same method.
Closing the account before exporting
Records only leave through the API, and analytics go with the account (7 days on Free, 30 on Grow, 90 on Grow Plus). Pull your no-results queries before you cancel.
Pasting the settings JSON as is
Each engine has its own vocabulary. An untranslated setting gets ignored without an error, and relevance drops with nobody knowing why.
Forgetting what writes to the index
A plugin, webhook or automation keeps sending new products to Algolia. The old index stays fresh, the new one goes stale. List every writer during inventory.
Undersizing RAM
Typesense keeps the index in memory. A cluster that's too small falls over on the first big import. Start generous, then scale down.
Comparing prices with different units
Algolia bills requests and records, Meilisearch a plan or a machine, Typesense server hours, Elastic VCU-hours, Doofinder request tiers. Compare on your real traffic: searches a month, keystrokes included, and catalog size without replicas.
Leaving old code in the theme
An uninstalled search app can leave scripts behind in a store theme. Shoppers type, nothing happens. Run a test search on mobile and desktop right after cutover.
- Records, settings, rules and synonyms exported via the CLI
- No-results queries exported from analytics
- Every writer to Algolia listed and redirected
- Replicas replaced by query-time sorting
- Results compared on the top 50 queries
- API keys revoked after a week of monitoring
Algolia alternatives: frequently asked questions
What is the best free alternative to Algolia?
For a static site or docs, Pagefind: it's free, open source, and runs in the browser with no server. For a dynamic app, Typesense and Meilisearch are free to self-host, so you only pay for the box they run on. Algolia's own Free plan still works for side projects: 10,000 search requests and 50,000 records a month, with 7 days of analytics.
Why is Algolia so expensive?
Because it bills per search request, and search-as-you-type fires a request on every keystroke. One shopper typing "red sneakers" can burn ten. On Grow, every 1,000 requests past the included 10,000 cost $0.50, and $1.75 on Grow Plus. Quick fix while you plan the move: debounce the input by 200 to 300 ms and you cut a big share of those requests without anyone noticing.
Typesense vs Meilisearch: which one replaces Algolia better?
Typesense if you want to keep your InstantSearch frontend almost untouched, since its adapter reads Algolia-style facetFilters and numericFilters. Meilisearch if you want the gentlest learning curve and a predictable bill from $20 a month. Both self-host. Typesense keeps the whole index in RAM, so size memory to your data.
Can I export my records from the Algolia dashboard?
No. The dashboard exports an index's configuration (settings, rules, synonyms), but records only come out through the API or the CLI, with
algolia objects browse. Run it before you cancel: once your API keys are gone, so is your data.Is there an open-source Algolia?
Not from Algolia itself: it's SaaS only, with no self-hosted edition. The closest open-source engines are Typesense (GPL-3.0) and Meilisearch (MIT), both built around typo tolerance and instant results, the two things people liked about Algolia in the first place.
What doesn't migrate from Algolia?
Merchandising rules, AI features (AI Ranking, AI Synonyms, personalization) and your analytics history. Rules export as JSON, but no other engine reads that file as is, so you rebuild them. Export your no-results queries before you leave: it's the best synonym list you'll ever get.
Is Algolia free?
Partly. The Free plan gives you 10,000 search requests and 50,000 records a month, 10 rules per index and 7 days of analytics. Past that you need Grow, which starts at $0 and bills $0.50 per extra 1,000 requests. DocSearch is a separate free program for open-source project docs and technical blogs, on application.
How long does an Algolia migration take?
For one index and an InstantSearch frontend, plan a few days: a day for export and settings, one or two for the frontend, then one to two weeks running both engines. An ecommerce catalog with dozens of rules takes longer, because each rule is rebuilt by hand.
Get the next migration guide in your inbox
One tool to leave, one place to go, one migration plan. Nothing else.