v0.9 -- 38 Cognitive Modules

38 Cognitive
Modules.
One Brain.

A neuroscience-inspired AGI framework implementing Free Energy Principle, Active Inference, and Global Workspace Theory. Runs locally. Zero API cost.

Try REPL

Requires Python 3.9+ • For CLI chat: also install Ollama

0
Cognitive Modules
0
Tests Passing
$0
Zero API Cost
MIT
Open Source License

The Problem

Current AI systems are narrow and disconnected. GPT-4 excels at language but lacks causal reasoning. Traditional cognitive architectures have limited modules. No system integrates all aspects of human-like cognition into a unified, locally-runnable framework.

Our Solution

ELO-AGI unifies 38 cognitive modules into a coherent brain-like architecture. Implementing Free Energy Principle, Active Inference, Global Workspace Theory, and Dual-Process Thinking -- all running locally with zero API costs under MIT license.

Module Explorer

38 Cognitive Modules

Each module implements a neuroscience-grounded capability. Click any tile to learn more.

View All Module Details

Evaluation

Benchmark Performance

Internal consistency metrics across 5 evaluation categories. Full methodology

Category Score Tests Status

How ELO-AGI Compares

Feature-level comparison against established cognitive architectures. Not a benchmark comparison.

Feature ELO-AGI ACT-R SOAR GPT-4
Modules 38 ✓ ~12 ~8 Monolithic
Local Execution Yes ✓ Yes Yes No
Causal Reasoning Native ✓ Rule-based Rule-based Prompted
Continual Learning Yes ✓ Limited Limited No
Zero API Cost Yes ✓ Yes Yes No
Open Source MIT ✓ Academic Academic No

Live Cognitive Processing

Watch the Brain Think

See how 38 modules coordinate to process information, from perception to reasoning to response.

View Full Demo

What You Can Build

Use Cases

Cognitive Research

Implement and test theories from Global Workspace Theory, Free Energy Principle, and Dual-Process cognition with modular, swappable components.

Causal Reasoning

Build applications with differentiable structural causal models, counterfactual reasoning, and causal discovery for decision support systems.

Local-First AI

Run all 38 cognitive modules locally with zero API cost. Full privacy, offline-capable, with modular design for easy ablation studies and prototyping.

Built with

Development Roadmap

Where We're Headed

v0.9 Current

38 cognitive modules, causal reasoning, robustness, hierarchical planning, PyPI distribution, REST API, sandboxed REPL.

1.0
v1.0 Next

External benchmark validation, real-time streaming responses, VS Code extension, test coverage gates.

1.1
v1.1 Planned

Multi-agent collaboration, voice interface, advanced tool integration, community plugin system.

Open Source, Community Funded

Sponsor ELO-AGI

Free and open source forever. Sponsorships fund research, development, and community growth.

All prices in EUR (€)

Individual Sponsor
€15 /month
  • Name in README
  • Early access to features
  • Priority GitHub Discussions access
Learn More
Organization Popular
€79 /month

Everything in Individual, plus:

  • Logo on README + website
  • Priority support channel
  • Quarterly strategy calls
  • Module integration guidance
  • Early access to new modules
Get Started
Research Partner
€299 /month
  • Co-development access
  • Custom module development
  • Dedicated support channel
Learn More

FAQ

Frequently Asked Questions

What is ELO-AGI and how is it different from GPT-4?

ELO-AGI is a neuroscience-inspired cognitive architecture with 38 specialized modules, not a single monolithic language model. Unlike GPT-4 which excels at language but lacks native causal reasoning, continual learning, and emotional processing, ELO-AGI implements these as dedicated cognitive modules that work together like brain regions. It runs entirely locally with zero API costs.

Do I need a GPU or special hardware to run ELO-AGI?

No. ELO-AGI runs on standard hardware with Python 3.9+. A GPU is optional and only needed for large-scale distributed processing via the neuro-scale module. Most cognitive modules run efficiently on CPU, including all reasoning, memory, and decision-making systems.

What does "neuroscience-inspired" actually mean?

Each module implements algorithms grounded in neuroscience research. For example: the Global Workspace module implements Baars' Global Workspace Theory of consciousness; the Dual-Process module implements Kahneman's System 1/System 2 framework; the Predictive Coding module implements Karl Friston's Free Energy Principle. These are not metaphors -- they are computational implementations of established neuroscience theories.

Is this actually AGI?

ELO-AGI is an AGI framework -- a comprehensive cognitive architecture designed to exhibit general intelligence capabilities. While no system today achieves human-level AGI, ELO-AGI addresses critical gaps identified in recent research: causal reasoning, compositional generalization, continual learning, and robustness. All 627 tests pass across these categories, demonstrating meaningful capability.

How can I contribute or get involved?

ELO-AGI is fully open source under the MIT license. You can contribute on GitHub, report issues, submit pull requests, or become a sponsor. For research collaboration, check the Research Partner tier on our pricing page. Contact: elvizekaj@zedigital.tech