01Memory
Memory & continual learning
Persistent memory that updates, merges, contradicts, and forgets as your system experiences the world.
The memory cloud for living machines.
Ailuruth gives agents persistent memory, spatial context, user profiles, connectors, and real-time recall. Extremely low latency. Works with any model.
$npx ailuruth setupCopied
Used by early design partners
NorthLabMeridianKaviHelixAtelierObsidianFieldKitVectorNorthLabMeridianKaviHelixAtelierObsidianFieldKitVector
> PRODUCT CATALOG[1/9]
All the primitives to build memory into your agent.
Focused infrastructure for ingesting, understanding, routing, and recalling context.
02Graph
Spatial memory graph
People, places, events, and objects live in relation to one another, not as loose chunks in a namespace.
03Recall
Sub-300ms traversal
Memory is fast enough to sit inside the request loop for live agents, companions, and embodied systems.
04Connectors
Every source, one memory
Conversation, documents, sensors, and app events continuously flow into the same living structure.
05Profiles
User and world profiles
Ailuruth keeps coherent profiles for users, rooms, machines, projects, and long-running jobs.
06Extract
Experience extraction
Raw context is parsed into facts, changes, locations, preferences, and traces that models can use.
> WHAT WE DO[2/9]
Bring your data. We build understanding. Your agent just knows.
Context infrastructure for AI systems. One API, every memory capability.
For developers & teams
The Ailuruth API
Memory, profiles, connectors, and graph traversal in one endpoint.
<300msrecall latency10B+memory tokensAnymodel
For everyone
Personal Ailuruth
One memory across assistants, tools, devices, and the work you repeat.
profile.mdroom.graphtoday.trace
Legacy RAG
Stores chunks. Returns chunks. Each session starts from zero.
- Embeddings, namespaces, spaces
- Append-only; no contradiction logic
- Retrieval, not memory
Ailuruth
Facts that evolve. Knowledge that merges, contradicts, and gets forgotten.
- User profiles and spatial graph
- Update, merge, contradict, infer
- One API: ingest, retrieve, remember
> HOW IT WORKS[3/9]
Install, ingest, understand, remember, retrieve.
Four primitives build the graph. The fifth makes it useful inside live thinking.
quickstart.ts
TSPYcURL
// 1. Install the SDK
npm install ailuruth
// 2. Initialize memory
import { Ailuruth } from "ailuruth";
const memory = new Ailuruth();
// 3. Remember an experience
await memory.remember({
user: "maya",
text: "The red toolbox moved to shelf two."
});
// 4. Recall inside the answer loop
const context = await memory.recall("toolbox location");Put Ailuruth between your model and the world.
Use the API with any model, agent framework, robot runtime, or product surface.
Stream in conversations, files, events, and sensors.
Context arrives continuously instead of waiting for a manual import or offline index.
Resolve entities as they change over time.
The graph learns who, what, where, and why as new evidence appears.
Memory forms inside the network.
Useful context becomes a living structure your agent can traverse in real time.
Recall becomes part of thinking.
Every request can reach into memory before the answer is formed.
> BENCHMARKS[4/9]
We do not think benchmarks tell the full story.
But memory has to be fast, accurate, and cheap enough to use every time.
94recall quality280msrecall time1 APImemory stack
Recall time
Memories returned in milliseconds, ready for live conversations.
AiluruthRAGLegacy
Feature coverage
Memory graphBestPartialNo
User profilesBestPartialNo
ConnectorsYesPartialNo
Sub-300ms latencyYesNoNo
> USE CASES[5/9]
Best for latency, quality, and context that changes.
AI companions that remember your users.
Personal assistants that remember the person across every session, tool, and device.
In production atKaviEndearNorthLab
Robots that remember the room.
Embodied agents that remember rooms, tasks, object locations, and operator preferences.
In production atFieldKitHelixOrbit
Realtime knowledge for agents.
Agents grounded in fresh project context, product docs, tickets, calls, and user state.
In production atVectorMeridianAtelier
Team memory that stays current.
Team knowledge that compounds instead of disappearing into chat history and document drift.
In production atObsidianNorthLabKavi
> ENTERPRISE[6/9]
Ailuruth runs everywhere.
Managed, inside your cloud, or fully local with the same memory model and the same API.
Managed cloud
One endpoint, elastic memory, and no graph infrastructure to operate.
Your VPC
Run inside your perimeter with your keys, network controls, and audit trail.
On device
Local-first recall for robots, edge agents, and sensitive offline environments.
SOC 2 readyGDPR compliantData stays in your perimeter
> TESTIMONIALS[7/9]
It finally feels like the model remembers, instead of pretending to.
We deleted half our retrieval stack the week we switched.
Our robot stopped re-learning the same building every morning.
40%fewer tokens
Compared with retrieval-only baselines in internal tests.
> PRICING[8/9]
Simple pricing, by usage. Pay only for what you use.
Every plan includes monthly credits and the same memory primitives.
Free
For experiments and prototypes.
$0 /mo
- 10k memories
- Shared infrastructure
- Community support
Pro
For teams shipping memory into products.
$49 /mo
- 1M memories
- Sub-300ms recall
- Email support
- All connectors
Scale
For production agents with heavy context.
$399 /mo
- Unlimited memories
- Priority recall
- Spend controls
- VPC option
Enterprise
For embodied fleets and regulated teams.
Custom
- On-device memory
- Dedicated support
- SSO
- Custom SLAs
> FAQ[9/9]
The fine print, in plain English.
A memory layer that lives inside the network rather than beside it: spatial, continual, and queried in real time as part of thinking.
For most memory use cases, yes. Retrieval fetches the past; Ailuruth grows from it. Teams can still keep their existing stores while moving memory into the graph.
Any model. Ailuruth sits between your model and your users, so it is model-agnostic and framework-agnostic.
Yes. Memory can run on-device for embodied systems so recall stays fast without a round trip.
You can run Ailuruth fully managed, inside your VPC, or on your own hardware. Your keys, your perimeter.
Build minds that remember.
If this is the kind of system you want to build, come build it with us.