CogNexus is a personal knowledge management platform with AI Agent capabilities. It uses a three-layer architecture (Facts, Relations, Abstractions) to store fragmented knowledge and discover hidden connections.

What LLM providers are supported?

CogNexus supports multiple LLM providers including Anthropic (Claude), OpenAI (GPT), DeepSeek, and Doubao (豆包). Each Agent can have its own LLM configuration.

Is Wielding.ai open source?

Yes. CogNexus is an open-source personal knowledge management and AI Agent platform under the MIT License, available on GitHub.

Personal Knowledge AI Agent

Q: What is the three-layer knowledge architecture?

The three layers are: Fact Layer (raw knowledge fragments stored in SQLite + Qdrant), Relation Layer (connections between facts discovered via Neo4j knowledge graph), and Abstraction Layer (higher-level insights auto-generated from patterns in your knowledge).

Q: What are Character Agents?

Character Agents let you create AI personas (like Laozi or a domain expert) grounded in a curated knowledge base. You can chat with them and they respond from their unique perspective.

Three-layer knowledge architecture to capture fragmented insights, discover hidden connections, and create AI character agents — your personal cognitive companion.

Browse Agents

📝

Capture Knowledge

Store facts, ideas, and insights into your three-layer knowledge base anytime

🔗

Discover Relations

AI automatically finds hidden connections between your knowledge fragments

🎭

Create Characters

Build AI personas grounded in curated knowledge — chat as Laozi, Feynman, or your own expert

🌐

Share & Explore

Publish your Agent, let others chat with your knowledge base

Why CogNexus?

Most note-taking tools store information but never help you think. CogNexus is different — it uses a three-layer architecture to not only store your knowledge fragments, but actively discover hidden connections between them and generate higher-level abstractions. Combined with AI character agents, it becomes your personal cognitive companion.

"Knowledge is not a collection of facts, but a web of connections." — The design philosophy behind CogNexus's three-layer architecture

From daily notes to deep research, from personal insights to character-based dialogues — CogNexus gives your knowledge structure, meaning, and an AI voice.

Core Features

📊

Three-Layer Knowledge

Fact Layer stores raw knowledge fragments with semantic search (SQLite + Qdrant). Relation Layer discovers connections via knowledge graph (Neo4j). Abstraction Layer auto-generates higher-level insights from patterns in your knowledge. Use /research, /today, /abstract commands to interact.

🎭

Character Agents

Create AI personas grounded in curated knowledge bases — a philosopher's writings, a scientist's methodology, or a domain expert's insights. Configure persona, chat in character, and let them auto-research based on their identity. Each character has its own namespace and LLM configuration.

🌐

Visualization & Sharing

Visualize your knowledge with force-directed Graph view, hierarchical Tree view (abstraction layer), and Timeline. Publish your Agent and let others chat — pay-per-conversation with ATP. Supports multiple LLM providers: Anthropic, OpenAI, DeepSeek, and Doubao.

🧠

CogNexus Platform

Personal Knowledge Management & AI Agent Platform

CogNexus uses a three-layer knowledge architecture — SQLite for metadata, Qdrant for bilingual semantic vector search (BGE-M3), and Neo4j for knowledge graph and relation discovery. Supports multi-namespace management, per-agent LLM configuration (Anthropic / OpenAI / DeepSeek / Doubao), and rich visualization with Graph, Tree, Timeline, and Chat views.

📊 SQLite + Qdrant + Neo4j 🔍 Semantic Search 🕸️ Knowledge Graph 🎨 Visualization 🤖 Multi-LLM

Total Agents

Knowledge Nodes

Registered Users

Frequently Asked Questions

What is the three-layer knowledge architecture? +

The three layers are: Fact Layer (raw knowledge fragments with semantic search via SQLite + Qdrant), Relation Layer (connections between facts discovered through Neo4j knowledge graph), and Abstraction Layer (higher-level insights auto-generated from patterns). This architecture lets you store fragmented knowledge and progressively build understanding.

What are Character Agents? +

Character Agents are AI personas built on curated knowledge bases. You can create characters like Laozi, Feynman, or a domain expert, configure their persona and background, then chat with them. They respond from their unique perspective, grounded in the knowledge you've provided. Each character has its own namespace and LLM configuration.

Which LLM providers are supported? +

CogNexus supports Anthropic (Claude), OpenAI (GPT), DeepSeek, and Doubao. Each Agent can be configured with its own LLM provider and model. You can set a global default in Settings and override per-Agent as needed.

How does Agent monetization work? +

Publish your Agent and set a per-conversation ATP price. Other users spend ATP to chat with your Agent, and you earn ATP revenue automatically. New users receive 100 ATP on registration. You can also set your Agent as free to maximize reach.

Is CogNexus open source? +

Yes. CogNexus is fully open-source under the MIT License. You can self-host and customize it freely. The source code is available on GitHub, and contributions are welcome.