Meilisearch

What Is Meilisearch?

A developer at a Paris-based e-commerce startup needed site search that returned results as users typed. Algolia was the obvious choice — until the pricing landed. At meaningful scale, Algolia's per-search billing produces invoices that can exceed the cost of the engineering team that built the feature. Elasticsearch was free but required a dedicated search engineer to tune, configure, and operate it. Neither option felt right for a ten-person team shipping product fast.

That tension is exactly what Meilisearch was built to resolve. Founded in 2018 by Thomas Payet and Guillaume Mourier in Paris, the project started as an open-source search engine written in Rust — chosen for memory safety and raw throughput rather than familiarity. The design philosophy was aggressive: sensible defaults that make search work without configuration, a RESTful API that any developer can use in a day, and a single binary that deploys in minutes. The goal was Algolia-quality results at the operational cost of a self-hosted tool.

In 2021 Meilisearch raised $15 million Series A from Felicis, CRV, and Seedcamp to build the managed cloud offering. Today Meilisearch Cloud adds EU-hosted infrastructure, a search analytics dashboard, and usage-based billing that starts at $30 per month — accessible for projects that need managed infrastructure without building their own.

The open-source core remains MIT-licenced. Embedding it in a commercial product, redistributing it, or modifying it without publishing changes is all permitted. For a French company operating under EU law, the combination of open licensing and EU data residency covers most compliance requirements without special negotiation.

Key Features

Typo Tolerance and Instant Results by Default

Most search engines require explicit configuration before typo correction works. Meilisearch enables it by default with a two-character edit distance that catches the most common misspellings — "recieve" finds "receive", "calender" finds "calendar" — without configuration files. Results return in under 50 milliseconds for typical document sets, making search-as-you-type feel immediate.

Ranking is also pre-configured: results order by relevance, then recency, then custom rules you define. For most applications, the defaults produce good results before any tuning. This matters for teams with limited search expertise; they get a working product on day one rather than a configuration project.

Hybrid Search: Keyword and Semantic in One Query

Meilisearch's most significant recent addition is hybrid search — the ability to run full-text and semantic vector search simultaneously in a single query, then merge the results using a configurable scoring ratio. The practical effect is that searching for "comfortable running shoes for bad knees" returns results even when no document contains those exact words, because the semantic component understands intent rather than matching strings.

Embeddings can be generated automatically. Configure an OpenAI or Ollama embedder and Meilisearch handles the vector generation pipeline — no separate embedding service, no preprocessing step, no embedding database to maintain. For teams building retrieval-augmented generation (RAG) pipelines, this makes Meilisearch a plausible vector store that also handles traditional keyword search.

The semantic ratio is tunable per query. Set it to 0.0 for pure keyword matching, 1.0 for pure semantic, or 0.5 for equal weighting. This granularity is rare in search engines at this price point.

Filtering, Faceting, and Geo-Search

Beyond basic search, Meilisearch provides the faceting infrastructure that product catalogues and content directories require. Define filterable attributes and users can narrow results by category, price range, rating, or any other structured field. Geo-search lets you order results by proximity to a location or filter to results within a radius — useful for listings platforms, local search, and maps integrations.

These features use the same REST API as basic search, with no additional service or index required. The configuration is declarative: define your filterable and sortable attributes in the index settings and Meilisearch optimises accordingly.

Developer Experience

Meilisearch ships official SDKs for JavaScript, Python, Go, PHP, Ruby, Rust, and several other languages. The API surface is small — you can implement basic search in a few dozen lines of code. The documentation is comprehensive and provides working examples for common frameworks including Laravel, Strapi, and Next.js.

For teams using Laravel, the official Laravel Scout driver makes integration a configuration change rather than a development project. Strapi, a popular EU-built headless CMS, ships Meilisearch as a first-class search provider. The Meilisearch ecosystem has grown substantially since the Series A, with the company maintaining quality across its SDK portfolio.

Self-Hosting Simplicity

Download the binary for your platform, run it, and search is operational. There is no JVM to configure, no cluster to initialise, no Elasticsearch Yellow status to investigate. The Docker image is under 50MB. For local development, this means every developer on a team can run search locally without infrastructure overhead.

The operational simplicity carries a trade-off: Meilisearch indexes data in memory. A 1GB index requires roughly 1GB of available RAM. This is not a concern for most applications, but it means billion-document deployments require either significant hardware investment or a move to Elasticsearch, which handles disk-resident indexes more efficiently.

Pricing

Meilisearch's self-hosted open-source edition costs nothing and has no feature restrictions. For teams with the infrastructure to run it, this is the most cost-effective option at any scale.

Meilisearch Cloud starts at $30 per month on the Build plan, which includes 50,000 searches and 100,000 documents. Additional searches cost $0.40 per 1,000 over the included amount. For context, Algolia's equivalent tier costs significantly more for the same search volume, and Algolia is a US company without EU-native data residency.

The Pro plan at $300 per month includes 250,000 searches and 1,000,000 documents, plus the analytics dashboard — a meaningful addition for product teams that want to see which queries are failing to return results. Enterprise plans are custom-priced and add dedicated infrastructure, SSO, and premier support with guaranteed SLAs.

For startups and growing applications, the Build plan offers genuine value. For established products, the self-hosted option with commercial support contracts from the community is often the most economical path.

EU Compliance & Privacy

Meilisearch is developed by Meili SAS, incorporated under French law at 52 Boulevard de Sébastopol, Paris. French companies are directly subject to GDPR, and Meilisearch explicitly acknowledges compliance with Regulation 2016/679 in their privacy policy. Meilisearch Cloud hosts data in EU infrastructure, with no US-based data transfers required.

The self-hosted option is the strongest compliance posture: data never leaves your environment. For organisations in financial services, healthcare, or public administration, self-hosting on EU infrastructure with Meilisearch's MIT-licenced code provides full audit rights over the software and complete control over data flows.

The company does not use customer data for model training or product improvement. There is no telemetry in the self-hosted binary. For EU organisations evaluating search infrastructure, the combination of French HQ, EU hosting, and MIT licence addresses most procurement requirements without legal review.

Who It's Best For

SaaS product teams adding search to their application who need good results fast. The defaults work, the API is intuitive, and the time from zero to deployed search is measured in hours rather than weeks.

E-commerce platforms requiring faceted search, filtering, and instant results across product catalogues in the tens of thousands. Meilisearch handles this tier of scale comfortably with hardware costs well below Algolia's monthly bill.

EU-regulated organisations that need search with EU data residency and full audit rights. The MIT licence and French HQ combination satisfies most compliance requirements that a US-based Algolia cannot.

Teams building AI search features who want hybrid keyword-plus-semantic results without building a separate vector pipeline. Meilisearch's automatic embedding generation removes a significant infrastructure burden.

The Verdict

Meilisearch has grown from a fast open-source search experiment into a credible production search platform with genuine AI capabilities. The ease-of-use advantage over Elasticsearch is real. The cost advantage over Algolia is substantial. The trade-offs — RAM-bounded scale, limited enterprise controls — are honest limitations rather than oversights. For teams that fit the profile, Meilisearch delivers Algolia-quality results at a fraction of the price, built by a French company whose data handling you can inspect in the source code.

Frequently Asked Questions

Is Meilisearch GDPR compliant?

Yes. Meilisearch is developed by Meili SAS, a French company under French and EU law. Meilisearch Cloud hosts data in the EU. The self-hosted option keeps all data within your own infrastructure. The company explicitly complies with GDPR regulation 2016/679.

Can I self-host Meilisearch?

Yes. Meilisearch is MIT-licenced and can be self-hosted on any Linux, macOS, or Windows server. Download the binary, run it, and search is operational. Official Docker images are available. Self-hosting has no feature restrictions compared to the Cloud offering, except the analytics dashboard.

How does Meilisearch compare to Algolia?

Meilisearch offers comparable search quality at substantially lower cost, particularly for self-hosted deployments. Algolia has stronger analytics, more mature enterprise SLAs, and broader third-party integrations. Meilisearch's open-source model provides code inspection rights and self-hosting freedom that Algolia does not offer.

What is hybrid search and how does it work?

Hybrid search runs full-text (BM25 keyword) and semantic (vector) search in parallel on the same query, then merges results using a configurable scoring ratio. You control the balance between keyword precision and semantic relevance per query. Embeddings are generated automatically using OpenAI, Ollama, or HuggingFace models you configure.

Where is Meilisearch Cloud data hosted?

Meilisearch Cloud hosts data in EU infrastructure. The company is French and subject to EU data protection law. For strict data residency requirements, self-hosting on your own EU infrastructure provides complete control over data storage and processing location.