Algolia

What is Algolia?

Algolia is a hosted site search API and discovery platform: you send it records such as products, articles or help pages, and it returns ranked results for search boxes, category pages, recommendation widgets and AI assistants. The company now markets it as an AI search platform, and says more than 18,000 customers in over 150 countries use it for agentic, generative and keyword search experiences.

A bylined July 2021 report in the independent tech press described Algolia as best known for search-as-a-service, a developer-friendly API for adding real-time search to an app or website, and noted that the company was originally founded in France. The same report covered a $150 million Series D round at a $2.25 billion post-money valuation.

Three things shape most buying decisions:

• Hosted only. On-premises deployment is not currently offered, so your index lives on Algolia's infrastructure.
• Usage-based billing. Self-serve plans meter search requests and stored records, so search-as-you-type designs raise costs directly.
• Tiered AI. Keyword search starts on the Free plan and AI relevance tools on Grow Plus, while NeuralSearch, Smart Groups and AI Collections come with Elevate, which requires an annual contract.

Core features

Search and relevance

• Hybrid keyword and vector search: NeuralSearch merges keyword and vector results in milliseconds, and teams can change the NeuralSearch and keyword ranking weights for individual results.
• Natural-language search bar: an AI mode lets shoppers type questions in plain language while AI interprets intent to guide discovery.
• Language handling: the natural language processing toolset provides default support for all languages and advanced support for more than 70 languages.
• Experimentation: built-in A/B testing lets teams tune and test search, measure outcomes and launch new experiments in minutes.
• Autocomplete: an open-source JavaScript Autocomplete library builds suggest-as-you-type experiences and can also query data sources other than Algolia.

Merchandising, browse and recommendations

Merchandising tools let teams adjust how products appear across category and collection pages, pinning items, promoting campaigns or changing ranking strategies without engineering support. Browse applies personalization, recommendations, rules and ranking to category and product listing pages.

The Recommend API covers trending items, trending facets, related products, frequently bought together and "looking similar" models, among other recommendation types.

Generative and agentic features

• Grounded answers: Algolia positions its retrieval-augmented generation (RAG) as grounding AI responses in your verified data while staying consistent with your search logic.
• Agent Studio: connects the LLM you choose to Algolia search and tools, manages the end-to-end workflow and grounds responses in live index data; completions are cached by default to reduce LLM token costs.
• MCP servers: three Algolia-managed Model Context Protocol servers cover AI-facing search (Public MCP), internal analysis (Productivity MCP) and API client snippets for coding agents (DevEx MCP).

Data ingestion

The Crawler is a low-code way to crawl a site and index it in Algolia; it can crawl HTML and non-HTML sources and password-protected pages, and crawls can be automated and scheduled. Teams that already have structured data can push records through the API clients instead.

Guide

The React quickstart builds an app that searches a sample product catalog in four steps:

1. Prepare an account and runtime. You need a free Algolia account and, for this sample, Node.js 20.19 or later.
2. Collect three credentials. In the dashboard's API Keys page, copy the Application ID, the Write API key (used by the indexing script to update records and settings) and the Search API key (used by the browser app to query the index).
3. Index and configure. An indexing script uploads products as records with saveObjects, then uses setSettings to make product_type a facet and to snippet the description. Keeping these as two steps lets you reapply settings without re-importing records.
4. Build the interface. React InstantSearch builds the search UI on top of the index, and the instantsearch.css package supplies base styles.

The quickstart is explicit about key handling: keep the Write API key secret, do not commit the .env.local file to version control, and never expose the write key in the app.

Algolia use cases and examples

• Ecommerce search and agentic commerce: beyond product search and category merchandising, Algolia pitches conversational and agentic experiences that guide shoppers to a confident purchase faster.
• Documentation search: DocSearch makes documentation sites and blogs searchable and is free for open-source and technical docs.
• Shopping assistants and conversational search: Agent Studio is pitched for shopping assistants that answer product questions and recommend items, content summarizers that extract insights from large datasets, conversational search, and custom workflows linking prompts to Algolia features.
• Customer-facing AI assistants: Public MCP exposes selected indices and Recommend models to external assistants for production uses such as product discovery and support bots.
• Internal search analytics: Productivity MCP answers business-intelligence-style questions and one-off checks, such as spotting trending searches or zero-result queries, within each user's own permissions.

Who is it for

Algolia's plan descriptions map onto distinct team types:

• Production apps that only need keyword search: Grow is described as a fit for a production-ready project that needs keyword search only, while Grow Plus adds AI capabilities and higher limits.
• Large catalogs and enterprises: Elevate adds NeuralSearch, an enhanced SLA, annual discounts and more indexes for larger federated catalogs. Its card also lists 99.99% availability, SSO access and enterprise-level support plans.
• Ecommerce business teams: Merchandising Studio is an interface optimized for e-commerce business teams, combining AI and manual controls to raise conversion rates.
• New projects and small stores: the Free plan is pitched at people building a new project, launching a startup, running a small online store or exploring the product.

It is a weaker fit if you need to run search on your own servers: on-premises and FedRAMP services are not something Algolia currently provides.

Platforms

• API clients: current major versions exist for C#, Dart, Go, Java, JavaScript, Kotlin, PHP, Python, Ruby, Scala and Swift.
• UI libraries: InstantSearch ships for Android, iOS, plain JavaScript (InstantSearch.js), React and Vue, and Algolia for Flutter adds core search helpers for Flutter apps.
• Commerce platforms: listed ecommerce integrations include Salesforce CC B2C, Shopify, Adobe Commerce and commercetools, and insights events can be captured through Segment and Google Tag Manager.
• Management tools: changes are made through the Algolia CLI, the web dashboard or an API client.
• AI assistants: Public MCP integrates with MCP-compatible platforms including ChatGPT, Claude, Vertex AI, Perplexity, Amazon Q, Microsoft Copilot and Salesforce Agentforce.
• LLM providers in Agent Studio: Anthropic, OpenAI, Azure OpenAI, Google Gemini, DeepSeek and OpenAI-compatible endpoints can be connected.
• Hosting: Free, Grow and Grow Plus run in US, UK and EU West locations; Elevate is listed as global.

Pricing

Algolia pricing combines a separate Free plan with two pay-as-you-go plans and an annual Elevate contract. Prices below are list figures as captured on October 5, 2026.

Grow and Grow Plus are billed monthly with no contract, while Elevate requires an annual contract. Beyond search and records, Grow and Grow Plus include 10,000 Recommend requests a month and then charge $0.60 per additional 1K, and crawling is included up to 10,000 crawls a month before $0.80 per additional 1K. The Free plan instead includes 5K recommendation requests and 5K crawls a month. On Grow and Grow Plus you are billed only for usage above each product's free tier, and only for products you choose to use.

How usage is counted

• Search requests: one request is a group of one or more search operations sent in a single exchange. In autocomplete and search-as-you-type interfaces, every keystroke is a new search request.
• Records: the total is the sum across all indices, and each pre-sorted copy of your records counts toward that total.
• Overage on records: going over the included amount is charged against the maximum number of records reached, but Algolia ignores the three days with the most records each month before calculating it.
• Estimator: the usage estimator on the pricing page gives starting prices that are directional only and not a guaranteed or binding quote.

Plan differences and AI costs

Grow covers keyword search and browse, query suggestions, 10 rules per index and manual synonyms. Grow Plus raises rules to 10,000 per index and adds AI Synonyms, AI Ranking, Advanced Personalization, Query Categorization and Collections. Volume discounts are available with an annual contract.

AI features carry costs outside the plan table. In Agent Studio you pay your LLM provider for tokens your agents consume, and search requests made by agents count toward your Algolia usage. Public MCP usage counts toward your existing plan, with no separate MCP fee.

Discount programs

• Startups: companies under three years old that have raised less than $5M can apply for $10,000 in credit, usable in the first 12 months after approval; unused credit lapses after 12 months, and existing paying customers are not eligible.
• Open source and non-profits: qualifying projects can use Algolia at a discount or for free, provided the project does not promote a product or service.

Algolia alternatives

Two open-source engines take a different approach to hosting and billing:

• Typesense: an open-source search engine you can run with Docker, native binaries or RPM/DEB packages, or provision through Typesense Cloud. Its cloud pricing is cluster-based, giving you a dedicated cluster with no limits on records or operations.
• Meilisearch: open source under the MIT license, so you can deploy it on your own infrastructure with no usage limits. Self-hosting is free but leaves updates, backups and scaling to you; Meilisearch Cloud offers usage-based or resource-based billing.

The trade-off is operational. Algolia meters requests and records and operates the infrastructure; self-hosted Typesense or Meilisearch avoids per-request charges in exchange for managing the servers yourself.

Limitations

Free plan terms

• Hitting the usage limit: once Free plan use reaches its limit, Algolia may suspend the service or apply an Algolia attribution to search results until you choose a plan that fits your usage.
• Size caps: if records exceed the Free plan's size limit or an application exceeds its data cap, Algolia will degrade or suspend service until usage fits the plan.
• Inactivity: if a Free plan account stays inactive for more than 30 days, Algolia may remove its data.
• Crawler cut-off: the free Crawler allowance is 5,000 resources a month, after which Crawler is suspended until you move to a suitable plan.
• Logo requirement: if you only use the free usage units, Algolia asks you to show its logo next to your search bar; mobile apps that cannot show it are asked to mention "Search powered by Algolia" in the store description and About page.
• Recommendations prerequisite: Recommend on the Free plan depends on implementing the Algolia Insights API.

Plan ceilings

Only Elevate allows more than one application, and the Free plan's 1 GB index and 10 KB record caps rise to 100 GB and a 100 KB record maximum on paid plans.

AI feature boundaries

• LLM access: Algolia provides free access to OpenAI's GPT-4.1 for testing Agent Studio, but production requires your own LLM provider account.
• Read-only MCP: Public MCP and Productivity MCP can search, recommend and report analytics, but they do not create or change indices, records or settings.

Privacy and data use

• Certifications and locations: Algolia states that its Search API complies with GDPR, CCPA, SOC 2, ISO 27001, ISO 27017 and C5, with processing locations across the EU, UK, USA, Canada, Brazil, the UAE, South Africa, India, Singapore, Hong Kong, Japan and Australia. Dashboard data is processed in the USA.
• Model training is on by default: all applications in an organization are automatically included in NeuralSearch model training, and application owners can change this in the dashboard.
• Training scope: Algolia says training currently uses anonymized data from Elevate customers only, excludes customers using HIPAA-enabled features, and works from a 90-day data retention window.
• Opting out: excluded applications are left out of future training runs, but models already trained stay unchanged.
• End-user requests: when Algolia processes personal data on a subscriber's instructions, end users are directed to contact that subscriber organization to exercise their rights.

FAQ

Q1. Is Algolia free to use?

Yes, within limits. The Free plan includes 10K search requests a month and 50K records with no credit card, but it asks for the Algolia logo and can be suspended or attributed once you reach the limit. Grow and Grow Plus also start at $0 a month and bill only usage above their included amounts.

Q2. Why can autocomplete raise my Algolia bill?

Each keystroke in a search-as-you-type interface counts as a separate search request, so instant search generates more billable requests than a submit-once form.

Q3. Does Algolia train AI models on my data?

NeuralSearch training includes applications by default, but Algolia says it currently uses anonymized data from Elevate customers only. Application owners can opt out in the dashboard; models already trained remain unchanged.

Q4. Can I self-host Algolia?

No. On-premises deployment is not currently offered. Open-source engines such as Typesense and Meilisearch can be self-hosted instead.