Algolia

A hosted search API for websites and apps: you push JSON records, and browsers query it directly for ranked, typo-tolerant results as the user types.

What it is: Search indexes
Type: Commercial product
Made by: Algolia
License: Proprietary
Facts checked: 24 October 2026

What it is

Algolia is a search engine you only reach over an API. You send records, small JSON objects with an objectID, into an index, choose which attributes are searchable and which can filter or facet, and query the index from your server or straight from the browser. The browser route is the usual one: a search-only API key is safe to ship in front-end code, and secured keys derived from it can limit each user to a subset of records. Every write is asynchronous. Algolia queues the operation and returns a taskID; until that task finishes, searches still see the old data.

The engine was built for search as you type: results have to come back at every keystroke, so it indexes word prefixes as well as whole words. Ranking doesn’t compute one blended score. It sorts by a list of criteria and uses each one only to break ties left by the ones before: number of typos, geo distance, filters matched, proximity of the query words, which attribute matched, exactness, and finally your own custom ranking such as popularity or sales. NeuralSearch, added in 2023, runs a vector search beside the keyword one and blends the two shortlists into one ranking.

Each application runs on a cluster of three identical servers in different data centres, kept in step with the Raft algorithm, and optional Distributed Search Network (DSN) replicas sit closer to users for reads. Pricing is by search requests and stored records: the Grow plan includes 10,000 searches and 100,000 records a month, then bills per extra thousand; NeuralSearch comes only with the custom-priced Elevate plan.

Why teams pick it

Nothing to run: no shards, nodes or upgrades, and a three-server cluster per application for availability
Built for instant search: prefix matching and typo tolerance on every keystroke, queried straight from the browser with a search-only key
Tie-breaking ranking that is easy to reason about: the order of the criteria, not tuned coefficients, decides which result wins
Front-end building blocks such as InstantSearch, and DocSearch, which is free for open-source and technical documentation

Watch out for

Closed source and hosted only; your index lives in Algolia’s infrastructure and you can’t run the engine yourself
Cost grows with every search request and record, so autocomplete that queries on each keystroke multiplies the bill
Records are capped at 10 KB to 100 KB depending on plan and an index at 100 GB; long documents have to be split into many records
Writes are asynchronous: a successful response means the change was accepted, not that it is searchable yet, and indexing pauses if one of the three cluster servers is down

Good for

Full-text search with relevance rankingWord prefixes are indexed for results at every keystroke, with typo tolerance and tie-breaking ranking on typos, proximity and exactness.

Specs

Compare with alternatives →

Data model: JSON records with an objectID, in indices
Query language: Search API with filters and facets
Client protocol: HTTP API, called from servers or straight from the browser
Secondary indexes: Searchable attributes indexed with word prefixes; attributes for filtering and faceting
Consistency: Writes are asynchronous; searchable once their taskID finishes
Replication: Three servers per application in separate data centres, kept in step with Raft
Multi-region: Distributed Search Network (DSN) read replicas closer to users
Ways to run it: Hosted only
Pricing model: Per search request and stored record; Grow includes 10,000 searches and 100,000 records a month

Alternatives

What all search indexes have in common →Compare side by side →

MeilisearchOpen sourceAlso good for full-text search with relevance rankingA search engine in one Rust binary with a REST API: typo-tolerant, search-as-you-type results with sensible defaults, plus vector and hybrid search.

ElasticsearchCommercialAlso good for full-text search with relevance rankingThe distributed search and analytics engine built on Apache Lucene: JSON documents in, full-text, vector and log queries out, over a REST API.

OpenSearchOpen sourceAlso good for full-text search with relevance rankingThe Apache 2.0 fork of Elasticsearch and Kibana 7.10.2 that AWS started in 2021, now run by a Linux Foundation project.

pg_trgmOpen sourceA PostgreSQL extension that splits text into three-letter pieces, so LIKE '%foo%', regex and typo-tolerant searches can use an index.UrsaDBOpen sourceCERT Polska’s byte-trigram index for malware collections: it narrows a YARA hunt over terabytes of files down to a few candidates.

How it fits with other technologies

What uses Algolia

Docsycan search the site with DocSearch fromAlgoliaDocumentation sites

Push records with the v5 JavaScript client, then search as the user types · ts

// npm install algoliasearch@5
import { algoliasearch } from 'algoliasearch'

// the admin key belongs on the server; browsers get a search-only key
const client = algoliasearch('APP_ID', 'ADMIN_API_KEY')

await client.saveObjects({
  indexName: 'products',
  objects: [
    { objectID: '1', name: 'Noise-cancelling headphones', brand: 'Acme', sales: 1200 },
    { objectID: '2', name: 'Headphone stand', brand: 'Deskly', sales: 300 },
  ],
  waitForTasks: true, // writes are queued; wait until they are searchable
})

// prefix matching: the partial word “headph” already finds both records
const { hits } = await client.searchSingleIndex({
  indexName: 'products',
  searchParams: { query: 'headph' },
})
console.log(hits.map((h) => h.name))

See it in a real stack

No architecture uses Algolia yet. Drop it into the playground and wire it up yourself.Open playground →

Where it came from2012–2026

Nicolas Dessaigne and Julien Lemoine, two search engineers who had worked together at Arisem and Exalead, started Algolia in Paris at the end of 2012. Their first product was an offline search engine SDK for mobile apps. It didn’t sell, but developers kept asking for an online version, and the company turned into a hosted search API, launched in beta in March 2013.

Algolia joined Y Combinator’s Winter 2014 batch and moved to the United States. Its pitch was instant search: results at each keystroke, tolerant of typos, for sites and apps with structured records such as products, rather than the long documents that classic engines and TF-IDF scoring were designed for. The engine was written from scratch rather than built on an existing open-source engine, with indexing and search in separate processes so that indexing could never slow searches down.

In September 2022 Algolia bought Search.io, an Australian vector search company, and in May 2023 launched NeuralSearch, which merges vector and keyword retrieval in the same index. By April 2026, when Stephen Lynch became chief executive, Algolia described itself as an AI search and retrieval platform serving more than 18,000 businesses and 1.75 trillion searches a year.

Why it’s built this way

Tie-breaking instead of one blended score: A single floating-point score mixes text relevance with popularity, so a popular item can beat a better match, and tuning the formula for some queries breaks others. Applying criteria in order keeps each decision explainable and holds for the whole catalogue as long as the data’s structure stays the same.
Index every prefix: Search as you type has to answer at each character. Building inverted lists for all prefixes of a word, stored in a sorted radix tree, makes a partial word as cheap to look up as a whole one.
Separate indexing and search processes: Both compete for CPU and disk. Running them as different processes with different scheduling priorities, and keeping search off the disk, lets heavy indexing run without slowing queries.
Hosted only, queried from the browser: The first offline SDK failed while developers asked for an online version. A hosted API with search-only keys removed the back-end hop and the servers to run.

Timeline

2012
Dessaigne and Lemoine start Algolia in Paris with an offline mobile search SDK
2013
Hosted search API launches in beta in March
2014
Joins Y Combinator’s Winter 2014 batch and moves to the United States
2022
Acquires Search.io, a vector search company, in September
2023
Launches NeuralSearch, vector and keyword search in one API, in May
2026
Stephen Lynch becomes CEO in April

Sources13 · facts checked 24 October 2026

Improve this page on GitHub