Manticore Search
Open-source search database for full-text, vector, hybrid and conversational search, queried over SQL or HTTP JSON.
At a Glance
About Manticore Search
Manticore Search is an open-source search database from Manticore Software Ltd, written in C++. It combines full-text, fuzzy, vector, hybrid and conversational search with real-time indexing, and is queried through SQL (MySQL protocol) or a JSON HTTP API. The latest GitHub release listed is 29.9.0, and the code is published under GPL-3.0.
What It Is
Manticore Search is a database built for search workloads that runs on your own servers. Developers create tables, insert documents and run queries with SQL or JSON, using any MySQL client or official client libraries. The project describes itself as an alternative to Elasticsearch, and it accepts Elasticsearch-compatible writes and works with tools such as Logstash and Kibana.
Search Capabilities
- Full-text search with over 20 operators, over 20 ranking factors, stemming, lemmatization, synonyms and Chinese segmentation.
- Fuzzy search, autocomplete, spelling correction, faceting, geo-spatial search and table joins.
- Vector search using KNN with HNSW, auto-embeddings with automatic chunking, and hybrid queries that combine full-text and vector retrieval.
- Conversational search that retrieves documents by KNN, keeps conversation history and uses an LLM provider to answer, via SQL CALL CHAT or the HTTP /search endpoint.
- Percolate tables and Kafka integration for stream filtering.
Storage, Scaling and Security
Tables can use row-wise storage or columnar storage through the Manticore Columnar Library, and a docstore option holds values without needing RAM. Sharded tables and Galera-based multi-master replication distribute data across nodes. Built-in authentication and authorization cover MySQL, HTTP/HTTPS and replication access. Backup is available through manticore-backup and the SQL BACKUP command, including S3-compatible storage.
Setup Path
A one-line installer supports Linux and macOS with Homebrew. Docker images, RPM and APT repositories, and Windows instructions are also documented. Official clients exist for PHP, Python, Python asyncio, JavaScript, TypeScript, Java, Go, .NET, Rust and Elixir. Data can be synced from MySQL, PostgreSQL, ODBC, MS SQL, XML, CSV and Kafka. The vendor also offers consulting, support and development services.
Community Discussions
Be the first to start a conversation about Manticore Search
Share your experience with Manticore Search, ask questions, or help others learn from your insights.
Pricing
Open Source (Self-hosted)
Manticore Search core is free to use and published under the GPL-3.0 (OSI-approved) license; run on your own servers.
- GNU General Public License v3.0 (GPL-3.0)
- Full-text, vector, hybrid and conversational search
- Query over SQL or HTTP on your own servers
- Self-hosted; no hosted plan priced in the evidence
Capabilities
Key Features
- Full-text search with 20+ operators and 20+ ranking factors
- Fuzzy search and autocomplete
- Vector (KNN/HNSW) and hybrid search
- Auto-embeddings with automatic chunking
- Conversational search with LLM-backed answers
- Row-wise and columnar storage
- Real-time inserts
- Faceted and geo-spatial search
- Percolate stream filtering and Kafka integration
- Sharded tables and multi-master replication
- Built-in authentication and authorization
- Backup and restore including S3-compatible storage
- SQL and JSON HTTP APIs with MySQL protocol support
- Elasticsearch-compatible writes
