3 October 2026 · Search & Data · 4 min read
Every "Algolia is expensive" thread online skips the one thing that actually matters: the real numbers. I pulled them straight off Algolia's pricing page this week. The free Build plan gives you 10,000 search requests and 50,000 records a month, no credit card. Past that, the Grow plan has no base fee — you pay $0.50 per extra 1,000 search requests et $0.40 per extra 1,000 records. That is a genuinely usable free tier and a genuinely pay-as-you-go paid tier. The reputation for being expensive comes from what happens once an app actually gets traffic.
Algolia doesn't publish a flat "starting price" the way a SaaS tool does, because Grow is metered on two axes at once — requests and records — and both can blow past the free allowance independently.
| Plan | Compris | Excédent |
|---|---|---|
| Build (free) | 10,000 search requests/mo · 50,000 records | n/a — hard cap |
| Grandir | 10,000 search requests/mo · 100,000 records | $0.50 / 1,000 extra requests · $0.40 / 1,000 extra records |
Both numbers confirmed on Algolia's official pricing page, checked 2 October 2026. Note Grow's included record count (100K) is double Build's (50K) even though the request allowance is identical — the free tier isn't a crippled trial, it is a real limit most side projects never hit.
Search requests are the variable that actually moves with traffic — every page load with a search box, every keystroke in instant-search, every filter click can be its own request. Here is the monthly Grow bill at three realistic volumes, record count held at a typical 150,000-item catalog:
| Search requests/mo | Request overage | Record overage (150K) | Total/mois |
|---|---|---|---|
| 50,000 (small SaaS) | $20.00 | $20.00 | $40.00 |
| 500,000 (growing app) | $245.00 | $20.00 | $265.00 |
| 2,000,000 (busy marketplace) | $995.00 | $20.00 | $1,015.00 |
The jump from $40 to $1,015 is an 25x bill increase for a 40x traffic increase — the unit economics barely bend, because there's no volume discount built into the published per-1K rate at this layer. A marketplace doing 2M monthly search requests is paying over $12,000/year just on the search layer, before records, recommendations or any add-on features.
I'm not writing this to tell you to rip Algolia out. Instant-search-as-you-type, typo tolerance, and relevance tuning that actually works out of the box are hard to build in-house, and engineering time to replicate it usually costs more than the overage bill. The honest alternative isn't "build your own" for most teams — it's self-hosted or differently-priced search infra: Meilisearch Cloud starts at $20/month on either usage-based (per document/search) or resource-based (instance + storage) billing, confirmed on their pricing page this week, with a 14-day free trial. Typesense Cloud takes a third approach entirely — dedicated per-hour clusters sized by RAM and vCPU with no per-request metering at all, but it doesn't publish a single dollar figure anywhere on its pricing page; you configure a cluster in their calculator or talk to sales to find out what it costs.
That opacity is itself a data point: Typesense's own page pitches free migration support "for those migrating from Algolia," which only makes sense if enough people are shopping around at Algolia's price point to be worth courting.
Under 10,000 requests a month, stay on Build — it's free and the limit is generous enough that most internal tools and early-stage products never touch it. Once you're paying Grow overage past a few hundred dollars a month, it's worth an afternoon getting a real Meilisearch Cloud quote at your actual document count and query volume, and requesting a Typesense cluster estimate for comparison — both specifically position themselves against Algolia's pricing, so the quotes tend to come back fast. Keep Algolia if the relevance tuning and instant-search UX are load-bearing for conversion; the per-request fee is cheap compared to a worse search experience costing you sales.
Vous êtes nouveau dans la tarification basée sur l'utilisation en général ? Commencez par le guides gratuits sur les coûts des API. For how a different metered-by-volume API quietly got worse over time, see CoinGecko's Basic tier value problem.
Rates verified against Algolia's and Meilisearch's official pricing pages, 2-3 October 2026. Typesense Cloud does not publish fixed rates. Reference estimates — confirm current pricing and your actual request/record mix before budgeting.