Soul File
Core personality, long-term goals, and identity anchors. Rarely changes — always informs every interaction.
Persistent · Identity
⚡ Cloudflare Edge · SQLite · Vectorize
A Memory That
Thinks Like You
HuMem.cloud is a sovereign, edge-native cognitive memory platform. Give your AI agents persistent, private, lightning-fast memory — with zero cloud egress costs and sub-millisecond local recall.
<1ms
Local Recall
95%
Cost Reduction
100%
Private
300+
Edge Locations
Four Layers of Cognitive Memory
HuMem organizes AI knowledge into a private temporal knowledge graph with four intelligent layers.
User File
Behavioral patterns, preferences, and interaction history. Learns how you work to serve you better over time.
Adaptive · Behavioral
Agents File
Task-specific memory per AI agent or workflow. Each agent remembers its own context, tools, and past outcomes.
Contextual · Per-Agent
Memory File
Episodic short-to-medium term session memory. Captures conversations, facts, and events with temporal anchors.
Episodic · Temporal
Why HuMem.cloud?
Sub-Millisecond Recall
SQLite runs directly inside your Durable Object. No cross-region round-trips. Retrieval happens at the speed of local memory.
Sovereign & Private
Every user gets their own isolated Durable Object. Zero cross-tenant data. Your memory stays yours — at the edge closest to you.
95% Cost Reduction
Eliminate centralized vector DB egress fees. SQLite on Durable Objects costs a fraction of Pinecone, Weaviate, or Chroma.
Temporal Knowledge Graphs
Facts have timestamps. Relationships evolve. HuMem tracks not just what the AI knows — but when it knew it.
Hybrid Retrieval
Combines vector similarity search (Vectorize + BGE-M3) with relational SQLite context for maximum recall accuracy.
Universal API Compatibility
Simple REST endpoints. Works with OpenAI, Claude, Gemini, LangChain, LlamaIndex, or any custom agent system.
Live API Playground
Connect to your edge memory in real-time. Enter a User ID to get started.
🔑 Session Authentication
Your User ID creates an isolated, private memory namespace at the edge.
No session loaded. Enter a User ID above to begin.
📥 Ingest Memory
Record facts, conversations, or experiences. The AI extracts entities and relationships automatically.
🔍 Recall Memory
Query the memory space. Hybrid retrieval combines vector similarity with relational context for maximum accuracy.
Frequently Asked Questions
Everything you need to know about HuMem.cloud's edge-native memory architecture.
What is HuMem.cloud?
HuMem.cloud is a sovereign, edge-native cognitive memory platform that gives users and AI agents a persistent memory system that is private, fast, and inexpensive to operate. Each user receives their own isolated Cloudflare Durable Object with a dedicated embedded SQLite database, enabling sub-millisecond local recall without cloud egress costs.
How does HuMem.cloud eliminate cloud costs?
Traditional vector databases charge per query and per gigabyte of data transfer. HuMem.cloud runs entirely on Cloudflare's serverless edge network using SQLite embedded inside Durable Objects. Reads and writes are local — there is no cross-region data transfer, eliminating egress fees entirely and reducing retrieval costs by over 95% compared to centralized cloud databases.
What are Soul, User, Agents, and Memory files?
HuMem organizes cognitive memory into four temporal layers:
- Soul File — Core personality traits, long-term goals, and identity anchors that rarely change.
- User File — Behavioral patterns, preferences, and interaction history.
- Agents File — Task-specific memory per AI agent or workflow.
- Memory File — Episodic short-to-medium term session memory.
Together these form a complete private temporal knowledge graph.
What is a temporal knowledge graph and why does it matter?
A temporal knowledge graph tracks how facts and relationships change over time. Rather than storing static snapshots, HuMem records when beliefs were formed, when they were updated, and when they expired. This allows AI agents to reason about context that happened yesterday vs. a year ago, producing far more accurate and relevant responses.
How fast is HuMem.cloud compared to traditional vector databases?
HuMem.cloud achieves sub-millisecond local recall via SQLite embedded in Cloudflare Durable Objects. Traditional vector database queries (e.g., Pinecone, Weaviate) involve a round-trip to a centralized server, typically adding 50–200ms of latency. HuMem's edge-native architecture puts the database physically close to the user, delivering 10–100× faster retrieval.
Is HuMem.cloud private and secure?
Yes. Each user's memory is completely isolated within a dedicated Cloudflare Durable Object. No user shares infrastructure with another. Data never leaves the edge node serving that user without explicit permission. HuMem is designed for sovereign, private-first AI memory — your data is yours.
What AI models and frameworks is HuMem compatible with?
HuMem.cloud exposes standard REST API endpoints (POST /memory for ingestion, GET /memory for retrieval) compatible with any backend, AI framework, or LLM pipeline. It integrates seamlessly with OpenAI, Anthropic Claude, Google Gemini, LangChain, LlamaIndex, and any custom agent system that can make HTTP requests.
Who built HuMem.cloud?
HuMem.cloud is a product of BotVibe AI LLC, founded and built in Clintwood, Virginia. It is part of the AntiGravity platform suite — a collection of sovereign, edge-native AI infrastructure tools designed for the next generation of intelligent applications.