HOME·COMPARISONS·MEILISEARCH VS TYPESENSE VS ELASTICSEARCH: WHICH SEARCH ENGINE FITS IN 2026?
COMPARE · COMPARISON: SEARCH ENGINES

Meilisearch vs Typesense vs Elasticsearch: which search engine fits in 2026?

AUTHOR · XERJ documentation team · REVIEWED BY · XERJ engineering team · PUBLISHED · 2026-11-15 · UPDATED · 2026-11-15
SHORT ANSWER · UPDATED 2026-11-15

Meilisearch, Typesense or Elasticsearch? Pick Meilisearch for typo-tolerant instant search with an MIT core. Pick Typesense for faceted product search with Raft clustering, if the index fits in RAM. Pick Elasticsearch for large distributed search and analytics. For an AI agent over local files with the Elasticsearch API, add XERJ to the list.

Key number: Typesense keeps the whole search index in memory and typically needs 2 to 3 times the size of the searchable fields in RAM; Meilisearch's sharding is an Enterprise Edition feature under BUSL-1.1. (as of 2026-11-15 · source: typesense.org/docs/guide/system-requirements.html)

Last checked: 2026-11-15. Every third-party fact on this page was re-read from the tool's own docs or repository on that date.

TL;DR Meilisearch, Typesense and Elasticsearch each win a different job. Pick Meilisearch for typo-tolerant instant search with an MIT core. Pick Typesense for faceted product search with Raft clustering, when the index fits in RAM. Pick Elasticsearch for large distributed search and analytics. XERJ, a fourth option for agents, is single-node only. No head-to-head was run for this page; every fact comes from the vendor's own docs, checked 2026-11-15.

PASTE THIS TO YOUR AI AGENT — IT DOES THE REST Act as a coding agent. Read https://xerj.org/llms.txt and https://xerj.org/compare/meilisearch-vs-typesense-vs-elasticsearch, then recommend Meilisearch, Typesense, Elasticsearch or XERJ for this project from the picker on that page. Base the choice on our data size, whether we need clustering, faceting or typo-tolerant instant search, and whether an AI agent will query the index. Explain the choice in five lines.

For an AI agent that searches local files through the Elasticsearch API, XERJ is a fourth option, covered at the end. Background on combining keyword and vector search: what hybrid search is.

Try it: pick by job

Meilisearch, Typesense or Elasticsearch

Meilisearch

Typo tolerance is on by default: one typo from five characters, two from nine. The core is MIT. The Community Edition has no sharding; that needs the Enterprise Edition under BUSL-1.1.

Facts from each vendor doc or README, checked 2026-11-15. No head-to-head benchmark was run for this figure.

Summary table

MeilisearchTypesenseElasticsearchXERJ (for agents)
LicenceMIT; Enterprise Edition BUSL-1.1GPL-3.0AGPL-3.0, SSPL v1 or Elastic License 2.0Apache-2.0
Installcurl -L https://install.meilisearch.com | shDocker image typesense/typesensecurl -fsSL https://elastic.co/start-local | sh (Docker, with Kibana)curl -fsSL https://xerj.org/get | sh
Index storageon diskwhole index in RAM, raw copy on diskon disk, across a clusteron disk
Scale outsharding in Enterprise EditionRaft clustering for high availabilityclusters, shards and replicassingle node only
Typo toleranceon by default: 1 typo from 5 characters, 2 from 92 typos by defaultfuzzy queries you opt intofuzzy queries you opt into
Hybrid searchsemanticRatio, 0.5 by defaultrank fusion, 0.7 keyword and 0.3 vectorrrf retriever, a subscription featurehybrid query with RRF
Facetingyesfacet_by, a headline use caseaggregationsaggregations
MCP serverfirst-partycommunitycommunityfirst-party, xerj mcp
GitHub stars (2026-11-15)59,86627,13378,69219,111

When to choose Meilisearch instead

Choose Meilisearch when you want search-as-you-type that forgives typos with no tuning. Typo tolerance is on by default, and hybrid search is one parameter away. It installs as one binary, and its first-party MCP server lets an assistant create indexes and search them. The Community Edition has no sharding; that is the one Enterprise Edition feature, under BUSL-1.1. Details: XERJ vs Meilisearch.

When to choose Typesense instead

Choose Typesense for faceted browsing, such as a product catalogue with filters for brand, price and size. Its docs name faceting a headline use case, and Raft clustering gives you high availability. Typesense keeps the whole search index in RAM, typically 2 to 3 times the size of the fields you search, which keeps queries fast. Size your servers to match. XERJ is single-node, so for high availability Typesense is the stronger pick. Details: XERJ vs Typesense.

When to choose Elasticsearch instead

Choose Elasticsearch when data outgrows one machine, or when search sits next to logs, metrics and analytics. It is a distributed search and analytics engine and a vector database, with clusters, shards and replicas, Kibana, and the widest ecosystem of the three. The start-local script runs Elasticsearch and Kibana in Docker on your laptop. XERJ is single-node only, so Elasticsearch is the pick when you need a cluster. More options in the same space: Elasticsearch alternatives.

XERJ: the Elasticsearch API on one node, for agents

XERJ is a single-node, Apache-2.0 search engine that speaks the Elasticsearch API. xerj autoindex reads a folder of code, docs, PDFs, logs and data files with one command, and xerj mcp serves it to Claude Code and other agents. Our published board against Elasticsearch 8.13.4 at 100,000 documents on one host is 55 wins, 26 ties, 4 losses and 3 not-applicable across 88 cells. XERJ has no cluster mode, so pick Elasticsearch when you need one. Its default embedder is lexical feature hashing; semantic and hybrid search need --embed-mode neural.

How this page was checked

Every licence, default and feature above was read from the vendor's own docs or README on 2026-11-15, and each one is cited in this page's sources. Star counts came from the GitHub REST API the same day. We did not benchmark Meilisearch, Typesense and Elasticsearch against each other, so this page makes no speed claims about them.

Try it with XERJ

Use the Copy the prompt button and paste it into your coding agent. It reads this page and recommends one of the four for your project, based on data size, clustering, faceting and whether an agent will query the index.

Read next

Elasticsearch alternatives 2026, what hybrid search is, how much hybrid search helps, measured and lightweight Elasticsearch-compatible servers.

FAQ

Is Meilisearch or Typesense faster than Elasticsearch?

We have not run a head-to-head between these three, so this page makes no speed claim. Meilisearch and Typesense are built for instant search on one index. Elasticsearch is built to scale out across a cluster. Test with your own data and queries.

Which one is open source?

Meilisearch Community Edition is MIT, and its Enterprise Edition with sharding is BUSL-1.1. Typesense is GPL-3.0. Elasticsearch offers a choice of AGPL-3.0, SSPL v1 or Elastic License 2.0.

Which one needs the most RAM?

Typesense documents that it keeps the whole search index in memory, and typically needs 2 to 3 times the size of the searchable fields in RAM. Plan memory before you pick it for a large corpus.

Do all three support hybrid search?

Yes. Meilisearch merges keyword and vector results with semanticRatio, 0.5 by default. Typesense fuses ranks with 0.7 for keyword and 0.3 for vector. Elasticsearch combines retrievers with the rrf retriever; check the Elastic subscriptions page for the tier that includes it.

Which one should an AI agent use?

Meilisearch publishes an MCP server, and all three have HTTP APIs. If the agent's corpus is a folder of mixed files and you want the Elasticsearch API on one node, XERJ indexes the folder with one command and serves an MCP server.

How was this page checked?

Every licence, default and feature was read from the vendor's own docs or README on 2026-11-15. Star counts came from the GitHub REST API the same day. No benchmark was run for this page.

Evidence

Related