By AICUBE Technology · Written in Rust
Your AI forgets everything.
QilbeeDB remembers.
A graph database with a built-in memory for AI agents. It keeps what happened, what it means, how it's done and what is true — stamped in time, recalled in under a millisecond. Your assistants stop starting from zero, every single conversation.
- memories written per second
- 100k+
- connections per second
- 50k+
- to recall a simple fact
- <1 ms
In plain words
A mind needs more than a spreadsheet.
Most databases store rows in tables. Minds don't work that way. A mind connects — people to places, causes to effects, moments to meaning. QilbeeDB stores information the way a mind does: as a living web of connections, stamped in time.
Inside, it works like a beehive. Every fact is a cell in the comb. Cells link into patterns, patterns consolidate into long-term memory, and whatever stops mattering is gracefully forgotten — the same way you don't remember every breakfast, but you never forget how to ride a bike.
Four kinds of memory
Everything the mind remembers, QilbeeDB keeps.
Human memory isn't one thing — it's four systems working together. QilbeeDB is the first graph database built around all of them.
Episodic
the diaryWhat happened. Every conversation, decision and event — in order, with full context.
Semantic
the encyclopediaWhat things mean. Concepts, entities and how they relate to one another.
Procedural
the muscle memoryHow things are done. Skills and routines your agent has learned and refined.
Factual
the ledgerWhat is true. Hard facts, indexed and ready to recall in under a millisecond.
And it forgets on purpose. Short-term memories consolidate into long-term ones automatically; what loses relevance fades away. That's not a limitation — that's how remembering actually works.
Bi-temporal by design
Two clocks. One truth.
Every memory in QilbeeDB carries two timestamps: when it happened, and when the system learned about it. It sounds subtle — it changes everything. Your AI can answer not just “what do we know?” but “what did we know last Tuesday?”. Rewind beliefs. Audit any decision. Correct the past without erasing it.
Written in Rust
Fast enough to feel like instinct.
Memory is only useful if recalling it is instant. QilbeeDB is written in Rust on a storage engine tuned for graphs, so remembering never becomes the bottleneck.
- nodes created per second
- 100,000+
- relationships per second
- 50,000+
- simple query response
- <1 ms
Plays well with others
Your tools already speak its language.
No lock-in, no exotic query language, no rewrite. If you've ever touched a graph database — or even if you haven't — you're minutes away from your first query.
- OpenCypher
The graph query language that reads almost like English.
- Bolt protocol
Neo4j-compatible drivers and tools connect out of the box.
- HTTP REST API
Talk to it from anything that can make a request.
- ACID transactions
All-or-nothing writes you can bet the business on.
- Security built in
JWT, API keys and role-based access control with granular permissions.
- Observability
Prometheus metrics and distributed tracing, ready for production.
Where it shines
One database. Many superpowers.
AI assistants that remember
Customer agents that recall every past interaction — and get better with each one.
Multi-agent teams
Dozens of agents coordinating through one shared, time-aware memory.
Recommendation engines
Suggestions built on how tastes connect, not just what was clicked.
Fraud detection
Spot suspicious patterns hiding in networks of transactions.
Knowledge graphs
Company knowledge organized the way people actually think about it.
Social network analysis
Map influence, communities and how information really travels.
Get started
Running in one command.
Pull the image, point any Neo4j driver at port 7687, and ask your first question. That's the whole setup.
docker run -d \
--name qilbeedb \
-p 7474:7474 \
-p 7687:7687 \
-v qilbeedb-data:/data \
qilbeedb/qilbeedb:latestgit clone https://github.com/aicubetechnology/qilbeeDB.git
cd qilbeeDB
cargo build --release
./target/release/qilbee-serverpip install qilbeedbHTTP API on port 7474 · Bolt protocol on port 7687
MATCH (alice:User {name: $alice})-[:KNOWS]->(common)<-[:KNOWS]-(bob:User {name: $bob})
RETURN common.nameQuestions & answers
What people ask first.
What is QilbeeDB?
QilbeeDB is a high-performance graph database written in Rust, designed for AI agent systems. Beyond storing data as a graph of nodes and relationships, it gives agents a real memory: episodic, semantic, procedural and factual — all tracked across two timelines (when things happened, and when they were learned).
Do I need to be a graph expert to use it?
No. If you can describe a relationship in a sentence — “Alice knows Bob” — you can write it in OpenCypher, the query language QilbeeDB speaks. It reads almost like English, and the docs walk you through your first queries in minutes.
Is it compatible with Neo4j tools?
Yes. QilbeeDB implements the Bolt protocol, so existing Neo4j drivers, libraries and tools connect to it out of the box. There is also an HTTP REST API for everything else.
What does “bi-temporal” mean?
Every piece of information carries two timestamps: event time (when it happened in the real world) and record time (when the database learned about it). This lets you ask what the system knew at any moment in the past, audit decisions, and correct history without erasing it.
Is QilbeeDB free to use?
QilbeeDB is source-available under the Business Source License 1.1: free to use inside your own products and projects, with restrictions on reselling it as a database service. On January 1, 2029 it converts to the Apache 2.0 open-source license. Commercial licenses are available from AICUBE Technology.
How fast is it?
The project reports 100,000+ node writes per second, 50,000+ relationship writes per second, and sub-millisecond responses on simple queries — powered by Rust and a RocksDB-based storage engine.