ekoDB is live. Start building for free
Why ekoDB
One database instead of a stack of services
Most teams stitch together a document store, a cache, a search engine, a vector database, and an auth layer, then spend their time keeping all of them in sync. ekoDB ships every one of those as native capabilities in a single Rust binary.
The result is less infrastructure to operate, lower latency because the data never leaves the process, and a feature set that is built in rather than bolted on. The benchmarks below put that against Redis, MongoDB, and Postgres.
What is ekoDB?
One governed platform for enterprise AI - data, retrieval, real-time APIs, and a built-in agent runtime.
Run AI agents right next to your data, governed by the same auth, encryption, and audit trail as every query, or run the platform as a fast managed database with built-in chat, RAG, and vector search.
EnterpriseAIstallswhendata,retrieval,andagentsliveinseparatesystemsthateachhavetobesecured,integrated,andoperated.ekoDBbringsthemintoonegovernedplatform:documentstorage,key-valueoperations,full-textsearch,vectorsearch,andabuilt-inagentruntime,soteamsgofrompilottoproductionwithoutassemblingastack.
Performance
Performance that holds up in production
Built in Rust for predictable latency under load, ekoDB delivers sub-millisecond reads and writes across every CRUD operation and outperformed some of the world's most widely deployed databases in industry-standard YCSB testing, even while carrying auth, encryption, and search on every request. Consolidation without compromise: ekoDB matches or beats the specialized tools it replaces, with governance on by default. Read more about ekoDB's remarkable performance here.
Pinned run v0.56.0, 1M records, 64 threads. Reads and durable writes measured under YCSB; ekoDB figures include auth, per-operation encryption, and search overhead on every request.
Benchmarks
Proven against the systems you already run
In industry-standard YCSB benchmarking, ekoDB delivered higher throughput than MongoDB and PostgreSQL on every workload. Its key-value engine outpaced Redis on every workload too, while carrying per-operation AES-256-GCM encryption and built-in search that Redis does not provide out of the box.*
*1M records, 64 threads. Document engines in durable mode with matching fsync; key-value in cache mode. Pinned run v0.56.0.
Relative throughput across YCSB workloads A, B, C, D, and F (average ops/sec, higher is better; the fastest engine in each group is set to 100). ekoDB numbers carry auth, encryption, and search overhead on every operation that the compared engines do not.
Documents / JSON / Row
Durable mode, all databases with matching fsync settings. ekoDB throughput includes full auth, encryption, search indexing, and real-time subscription overhead on every operation.
Key-Value
Cache mode, pure throughput. ekoDB's key-value engine beat Redis on every YCSB workload (1.1x to 1.3x higher throughput, at lower latency) while carrying AES-256-GCM encryption and built-in search indexing that Redis does not provide out of the box.
Built-in, not bolted on
Every feature below needs a separate product, extension, or service in a typical stack. In ekoDB they are native.
| Feature | ekoDB | PostgreSQL | MongoDB | MySQL | Redis |
|---|---|---|---|---|---|
| Built-in AuthApplication-level JWT tokens and API key authentication | SupportedJWT + API keys | PartialRoles only | PartialSCRAM only | PartialRoles only | PartialACL only |
| Per-op EncryptionEvery read and write is encrypted at the operation level | SupportedAES-GCM | Partialpgcrypto ext | PartialCSFLE client-side | PartialTDE / app-level | Not supportedTLS in-transit only |
| Real-time SubsLive data streaming via WebSocket subscriptions | SupportedWebSocket | PartialLISTEN/NOTIFY | SupportedChange Streams | Not supportedneeds external | SupportedPub/Sub |
| Cross-DB SyncSync data across different database engines in real time | SupportedRipple - Pg, Mongo, Redis | Not supportedsame-DB only | Not supportedsame-DB only | Not supportedsame-DB only | Not supportedsame-DB only |
| Key-Value StoreNative key-value operations alongside documents | SupportedBuilt-in | Partialhstore ext | Partialsingle-field docs | Not supportedtable workaround | SupportedBuilt-in |
| Document QueriesRich queries on flexible JSON documents | SupportedBuilt-in | SupportedJSONB | SupportedBuilt-in | PartialJSON columns | PartialRedisJSON module |
| Tunable DurabilityChoose durability vs speed globally or per operation | SupportedGlobal + per-op | Partialsynchronous_commit | SupportedWrite concern | Partialinnodb_flush | Partialappendfsync |
| Full-Text SearchSearch across text fields with ranking and stemming | SupportedBuilt-in | Supportedtsvector | PartialAtlas Search | SupportedFULLTEXT index | PartialRediSearch module |
| Vector SearchSimilarity search on embeddings for RAG and AI | SupportedBuilt-in | Partialpgvector ext | PartialAtlas only | Not supportedneeds Pinecone | PartialRediSearch module |
| FunctionsRun logic on the server, co-located with your data | SupportedServer-side | SupportedPL/pgSQL | PartialAtlas only | SupportedStored procs | PartialLua scripts |
| SWR CachingStale-while-revalidate responses built into the server | SupportedBuilt-in | Not supportedapp-level only | Not supportedapp-level only | Not supportedapp-level only | Partialmanual TTL |
| AI ChatQuery your data in natural language with built-in RAG | SupportedNative RAG | Not supportedneeds external | PartialAtlas only | Not supportedneeds external | Not supportedneeds external |
One system to secure, not a stack to assemble
Every capability above ships in a single ~50 MB service - no extensions, no bolt-on databases, no separate vendors to reconcile, no separate processes to deploy, secure, and audit.
What that looks like in practice
Fewer systems, less glue, lower latency
One system to run
Document storage, key-value operations, full-text search, vector search, real-time subscriptions, authentication, and per-operation encryption all live in the same database, so there are fewer moving parts to deploy, monitor, and reason about.
Lower latency by design
Because compute sits next to the data, queries and search avoid the extra network hops that a separate cache or search service would add on every request.
Built in, not bolted on
Capabilities that need extensions or extra services elsewhere ship natively in one binary that stays straightforward to deploy and operate as you scale.
Frequently asked questions
Is ekoDB a document, key-value, or vector database?
All of them. ekoDB is multi-model, with document storage, key-value operations, full-text search, and vector search in one engine.
Does it support transactions?
Yes. ekoDB has ACID transactions with configurable isolation levels, so you choose how strict you want consistency to be. Transactions buffer your changes and apply them atomically on commit, support savepoints, and let you read your own pending writes. Durability is tunable too, so you can trade latency for stronger guarantees per operation.
How does it compare to Redis, MongoDB, and Postgres?
In industry-standard YCSB benchmarking ekoDB outperformed across workloads, including faster reads and writes than MongoDB and Postgres, and higher throughput than Redis while also providing auth, encryption, and search.
Is ekoDB managed or self-hosted?
ekoDB is a managed cloud database. You deploy an instance per region and scale each deployment by machine tier.
Can it sync with databases I already run?
Yes. Ripple keeps data in sync across engines, including Postgres, MongoDB, and Redis.
Get started
One database, ready in minutes
Replace the stack with a single managed deployment. Document, key-value, search, vector, auth, and encryption, all built in.