A Rust-powered NoSQL database with ultra-low latency, hybrid in-memory and persistent storage, and built-in security — engineered for distributed systems that need flexibility without sacrificing performance.
See your database, not just commands. Meow is the native desktop workspace for exploring stores, managing records, and working with Montycat semantic search from one focused interface.
Use Montycat's onboard embedding models, or supply precomputed vectors from the model you already use. Either way, records and vectors live together in one engine and are searched through the same native vector index. Available in the semantic edition.
# Semantic search is ON by default
# in the Montycat Semantic
hits = await Products.semantic_search_get_values(
"Show all Bluetooth devices",
limit=5
)
# [{
# __value__: {
# "name": "Wireless Headphones"
# },
# __key__: "1283…",
# __score__: 0.82
# }]Write your data normally and Montycat embeds and indexes it automatically with a model running on your hardware.
Supply record and query vectors from OpenAI, your own model, an existing vector store, or a multimodal pipeline.
Fast kNN similarity search over either onboard embeddings or vectors you supply.
Your records and their vectors in one engine. No sync jobs, no drift, no second service to run.
Persistent, semantically searchable retrieval for grounded answers and long-term memory.
Your vector index stays on your hardware, without a managed vector database to operate or trust.
Client-observed persistent-storage results, separated by how the vectors are produced.
Intel i9-11900K · 8 cores · 32 GB RAM
Apple M1 · 8 GB RAM · BGE-small
Directional context only. Published competitor results used different hardware, workloads, recall targets, and client concurrency.
Montycat result: Intel i9-11900K, 32 GB RAM. Competitor figures: published source ↗
A decentralized architecture supported by self-serve infrastructure, purpose-built to replace monolithic data lakes and warehouses with domain-driven, federated data products.
Federated governance is enforced by the engine, not by convention: a platform team delegates scoped administrative capabilities per store and per keyspace, with storage and model guardrails, explicit denials, and an auditable policy history.
Explore Data Mesh governance
Every feature is designed to remove friction between your teams and your data — from the engine core to the developer experience.
Lightning-fast data processing and retrieval, up to 1,000,000 operations per second on a single node.
Seamlessly integrate and manage data across domains. Create keyspaces with flexible configurations and a unified view of your data landscape.
TLS, RBAC, secrets management, and tamper detection — production-grade security out of the box.
Memory safety, no garbage collection pauses, and predictable performance for the data path that matters most.
Intuitive client libraries for Python, TypeScript, JavaScript, Dart, and Rust — no SQL injection by design.
NoSQL flexibility with SQL-like capabilities: native foreign keys, optional strict schemas, and adaptable workloads.
Montycat unifies records, vectors, onboard embeddings, live updates, and data ownership. See how that integrated architecture compares with databases that assemble the same workflow from extensions, services, modules, or application code.
| Montycat | PostgreSQL | MySQL | MS SQL Server | Redis | MongoDB | |
|---|---|---|---|---|---|---|
| Primary Data Model | NoSQL records + vectors | Relational + JSON | Relational + JSON | Relational + JSON | Key/value data structures | Documents |
| Runtime Architecture | Rust; memory-safe, no GC | C process engine | C/C++ process engine | C/C++ process engine | C process engine | C++ process engine |
| Storage Model | Native in-memory + persistent keyspaces | Persistent, shared buffers | Persistent, buffer pool | Persistent, buffer pool | Memory-first, optional persistence | Persistent, cached working set |
| Vector Search | Built-in HNSW | pgvector extension | HeatWave Vector Store | Native; ANN index in preview | Query Engine / vector sets | Search process (mongot) |
| Embedding Pipeline | On-device, embed-on-write, field-selective | Application-managed | HeatWave GenAI pipeline | Configured model integration | Application-managed | Search pipeline + model endpoint |
| Records + Vectors | One engine, one write path | Same DB with extension | HeatWave service | Same database | Same server / module | mongod synchronized to mongot |
| Live Change Delivery | Native subscriptions | LISTEN/NOTIFY or CDC | Binlog / CDC | CDC / change tracking | Pub/Sub or Streams | Change streams |
| Schema Strategy | Schemaless or strict per keyspace | Strict relational schema | Strict relational schema | Strict relational schema | Data-structure based | Flexible + optional validation |
| Application API | Typed clients; structured commands | SQL + drivers / ORMs | SQL + drivers / ORMs | T-SQL + drivers / ORMs | Commands + clients | Document API + drivers |
| Data Ownership & Governance | Owner/store/keyspace policies built in | Roles, schemas, extensions | Roles and schemas | Roles and schemas | ACLs and key patterns | Roles and databases |
| Built-In Security | TLS, scoped ownership, policy grants/denials, secrets, tamper detection | TLS, roles, privileges | TLS, roles, privileges | TLS, roles, permissions | TLS, ACLs, key patterns | TLS, roles, privileges |
| Self-Hosted AI Stack | One binary; no embedding API | DB + extensions + model service | Vector AI features in HeatWave | DB + configured model service | DB + application model | mongod + mongot; model optional |
