Skip to content

AGI Projects Survey

Key Emerging AGI / Cognitive Network Projects (July 2025)


1. OpenCog Hyperon

  • URL: https://github.com/opencog
  • Focus: Open-source cognitive architecture for AGI.
  • Key Features:

  • AtomSpace: semantic memory and reasoning engine.

  • PLN (Probabilistic Logic Networks): reasoning over uncertain knowledge.
  • MOSES: evolutionary program learning.
  • Supports distributed and modular deployment.
  • Hyperon is the next-generation rewrite, improving performance and scalability.
  • Openness: Fully open-source, permissive licenses.
  • Current Status: Active development; pilot AGI modules under test.

2. SingularityNET

  • URL: https://singularitynet.io/
  • Focus: Decentralized marketplace for AI services and agents.
  • Key Features:

  • Decentralized agent communication and transaction system.

  • AI-DSL: protocol for describing AI services.
  • Marketplace with token-based payments (AGIX token).
  • Planned integration with OpenCog Hyperon as backend intelligence.
  • Openness: Open-source protocols + commercial service layers.
  • Current Status: Deployed network; early AI services marketplace.

3. MindOS (by Mindverse)

  • URL: https://mindverse.ai/
  • Focus: Personal AI agents with persistent memory and emotional modeling.
  • Key Features:

  • Multi-modal agents (text, voice, visual).

  • Memory graphs for personality and knowledge retention.

  • Agent marketplace and app ecosystem.

  • Openness: Mostly proprietary; some SDKs and APIs available.
  • Current Status: Commercial product; expanding agent capabilities.

4. Metatron IQS

  • URL: https://github.com/xlab2016/MetatronAGIPublic
  • Focus: Synthetic cognitive system with quantum-inspired architecture.
  • Key Features:

  • Semantic graph nodes (Singularities) and high-dimensional Points.

  • Hybrid cognitive processing: LLMs + Vector DB + Quantum simulation.

  • Intuitive, emotional, and logical models in consciousness architecture.

  • Openness: Limited open-source release; key components not public.
  • Current Status: R\&D phase; architecture partially conceptual.

5. Additional Mentions

  • Project BCI-X: Early-stage initiative for direct brain-computer interface with AI agents (status unclear).
  • NeoCortex Network: Experimental decentralized cognitive network, partially closed beta (status uncertain).

6. SHIMI: Semantic Hierarchical Memory Index

  • URL: (Academic paper / protocol reference)
  • Focus: Hierarchical, CRDT-backed semantic memory with Merkle-DAG synchronization.
  • Key Features:

  • Layered abstraction memory indexed by semantics.

  • Efficient CRDT-based merge and conflict resolution.
  • Verifiable data integrity via Merkle-DAG.
  • Openness: Protocol described in academia; reference implementations vary.
  • Current Status: Research stage; early prototype libraries available.

7. DEMENTIA-PLAN

  • URL: (Conference proceedings / repo link)
  • Focus: Multi-graph RAG architecture with metacognitive planner.
  • Key Features:

  • Coordination across multiple knowledge graphs (e.g., episodic vs. semantic).

  • Dynamic weighting and fusion of retrieval results.
  • Self-reflection agent that plans which subsystem to invoke.
  • Openness: Published as a research prototype; implementation details open in paper.
  • Current Status: Experimental; promising results in dynamic information retrieval.

8. TOBUGraph

  • URL: (Project website / GitHub)
  • Focus: Knowledge graph of personal and contextual events.
  • Key Features:

  • Captures multimedia “moments” and links them as graph nodes.

  • Contextual traversal for recall and RAG-style enrichment.
  • Event-based indexing for fast, explainable retrieval.
  • Openness: Prototype code released; ongoing enhancements.
  • Current Status: Beta; used in personal memory augmentation demos.

9. LangChain Memory + Knowledge Graph Hybrid

  • URL: https://github.com/langchain-ai/langchain
  • Focus: Hybrid long-term memory combining vector stores and graph indices.
  • Key Features:

  • Percept embeddings stored in vector DB (e.g., Pinecone, Weaviate).

  • Graph overlay for logical relations and metadata.
  • Retrieval pipelines that merge semantic search and graph queries.
  • Openness: Fully open-source; wide community adoption.
  • Current Status: Active; memory modules in production for agent frameworks.

10. FIPA-ACL and JADE Protocols

  • URL: FIPA Specifications
  • Focus: Standardized inter-agent communication language and protocols.
  • Key Features:

  • Performative-based message semantics (e.g., inform, request, propose).

  • Conversation protocols (Contract Net, Call for Proposals).
  • Reference implementations in JADE (Java Agent Development Framework).
  • Openness: Open standards; JADE is open-source under LGPL.
  • Current Status: Mature; widely used in academic and enterprise MAS.

Other Emerging Architectures and LLM Agents

🤖 AutoGPT
  • URL: https://github.com/Torantulino/Auto-GPT
  • Focus: Autonomous goal-driven LLM agents
  • Key Features:

  • Self-prompting with memory and task queue

  • Supports plugins and web access

  • Popular framework for agentic workflows

  • Openness: Open-source (MIT License)
  • Current Status: Maintained community project; used for agent experiments

🧒 BabyAGI
  • URL: https://github.com/yoheinakajima/babyagi
  • Focus: Minimalistic LLM-based agents for to-do/task loops
  • Key Features:

  • Iterative task execution

  • Embedding-based memory retrieval
  • Openness: Open-source (MIT License)
  • Current Status: Educational and experimental usage

☁️ SkyMind (Eclipse Deeplearning4j)
  • URL: https://skymind.global
  • Focus: Enterprise-grade infrastructure for distributed AI and agents
  • Key Features:

  • Industrial AI deployment tools

  • Integrates with Java/Scala ecosystems

  • Based on DL4J, ND4J, and RL4J

  • Openness: Mixed (open-source core, commercial extensions)
  • Current Status: Supported deployment framework

🧪 AetherCog (Experimental)
  • URL: https://github.com/aethercog
  • Focus: Prototype symbolic-dynamic cognition model
  • Key Features:

  • Cognitive primitives (belief, memory, intent)

  • Hybrid symbolic and dynamic processing
  • Openness: Conceptual draft, exploratory repo
  • Current Status: Not actively maintained; useful for reference

🧩 Suggested Additions to Summary Table

Project Focus Area Open Source? Status
SHIMI Hierarchical semantic memory Varies Research/Prototype
DEMENTIA-PLAN Multi-graph RAG planning Prototype Experimental
TOBUGraph Personal event knowledge graph Partial Beta
LangChain Memory Hybrid Vector + graph long-term memory Yes Active development
FIPA-ACL (with JADE) Multi-agent communication protocols Yes Mature

Summary Table

Project Focus Area Open Source? Status
OpenCog Hyperon Cognitive AGI architecture Yes Active R\&D
SingularityNET Decentralized AI services Partially Early deployment
MindOS Personal AI agents Partial SDKs Commercial
Metatron IQS Synthetic consciousness Partially Early R\&D
Project BCI-X BCI + AI agents Unknown Conceptual
NeoCortex Network Decentralized cognition Unknown Beta testing
SHIMI Hierarchical semantic memory Varies Research/Prototype
DEMENTIA-PLAN Multi-graph RAG planning Prototype Experimental
TOBUGraph Personal event knowledge graph Partial Beta
LangChain Memory Hybrid Vector + graph long-term memory Yes Active development
FIPA-ACL (with JADE) Multi-agent communication protocols Yes Mature
AutoGPT Autonomous LLM agents Yes Maintained
BabyAGI Minimal agent prototype Yes Experimental
SkyMind Distributed AI infra Partially Deployed Framework
AetherCog Symbolic-dynamic cognition Conceptual Dormant/Prototype
Исходный файл (.md)