Memory-Native Engine

The data engine
that remembers.

Durbl is a memory-native data engine for AI agents. Give your models persistent state, evolving context, and long-term recall.

main.py
1from durbl import Durbl
2
3client = Durbl(api_key="drbl_...")
4
5# Write a memory
6client.memory.write(
7 entity="agent/nexus",
8 content="User preference: dark mode",
9 importance=0.9,
10)
11
12# Build context for an LLM
13ctx = client.context.build(
14 entity="agent/nexus",
15 goal="Configure UI",
16)
17
18print(ctx.prompt)
THE PROBLEM

AI applications have no memory

Large language models can reason, write, and create — but they can't remember.Every conversation starts from scratch. Every agent forgets its past. The problem isn't the model. It's the infrastructure underneath.

No Persistence

Your AI starts from zero every session. Users repeat themselves. Preferences are lost. Context disappears between conversations.

Fragmented Infrastructure

Developers stitch together 5-7 different systems to give AI basic memory. The result is fragile, expensive, and never truly intelligent.

No Learning Loop

Your AI can't learn from experience. It repeats the same mistakes. It can't build on past interactions. Every conversation is isolated.

How It Works

A new kind of data engine

Traditional databases store rows and documents. AI needs memory that persists, state that evolves, context that's computed, and a lifecycle that reinforces what matters. Durbl is built on 6 core primitives.

MemoryObject

The atomic unit. Content with quality scores for importance, confidence, recency, and durability.

memory

StateObject

Versioned entity state with immutable history. Every change creates a new version with full temporal queries.

state

ContextFrame

Goal-aware context assembled from memories, state, and policies. Ready to pass directly to any LLM.

context

MemoryLink

Typed, weighted connections between memories. Builds a knowledge graph that grows with every interaction.

relationships

CognitiveEvent

The raw input. Messages, tool calls, agent outputs — everything enters through events and flows through the pipeline.

ingestion

PolicyObject

Rules for memory lifecycle. Retention, decay, access control — configurable per entity and memory type.

governance
CAPABILITIES

Everything your AI needs to remember

Durbl provides the persistent infrastructure layer your AI applications need to maintain state, build context, and learn from experience.

Persistent Memory

Every data point is a living memory with its own lifecycle, importance score, and decay curve.

Adaptive State

Evolving state management for agents. Every interaction updates the entity's state with immutable versioning.

Goal-Aware Context

Assembles perfect context frames by weighing relevance, recency, and specific agent objectives.

Real-Time Sync

Ultra-low latency memory propagation. Changes in one agent's memory are instantly available across the platform.

Lifecycle Policies

Define how memories live and die. Automatic summarization, merging, and retention based on your rules.

Multi-Agent Memory

Agents share a unified memory space while maintaining individual private context.

SDK

Simple, powerful API

Durbl abstracts the complexity of memory and state into a clean, high-performance Python SDK.

example.py
1client.memory.write(
2 entity="agent/nexus",
3 content="System preferences set to high-performance",
4 importance=0.92,
5 tags=["config", "perf"]
6)
USE CASES

Built for every AI application

AI Agents & Copilots

Agents that learn from experience, remember past decisions, and share knowledge with other agents.

Chatbots & Assistants

Assistants that remember users across conversations — preferences, history, and context.

AI-Powered Apps

Mobile and web apps with personalized AI experiences that improve over time.

SaaS Platforms

Add intelligent memory to your SaaS — per-user personalization, behavior tracking, and predictions.

Customer Support AI

Support agents that know the customer's history, past issues, and preferred resolution style.

AI Education

Tutors that adapt to each student's level, learning style, and progress over time.

What Durbl is not

Not a vector database with better branding
Not a wrapper over existing databases
Not a chat history store or RAG tool
Not an agent framework or LLM provider
Not an analytics dashboard — intelligence is built into the engine

Durbl is a new category: memory-native data infrastructure for the age of AI.

Transparent and self-hostable

Durbl's core engine is source-available — you can inspect the code and self-host for internal use. Durbl Cloud provides the managed experience with advanced features and dedicated support.

Source-Available (BSL License)
Self-hostable for internal use
Full code transparency
durbl-os / corepublic
1 def initialize_engine():
2 engine = MemoryEngine()
3 engine.connect()
4 return engine

Give your AI
a memory.

Get started in minutes. Build AI applications that learn, adapt, and remember.

Get Started Free
pip install durbl