This project defines a three‑layer AI governance stack: Execution for automation, Symbolic Codex for intent validation with algorithmic AI endorsement, and Normative Constitution for invariants and consensus via a second AI integration. Together they form a constitutional loop ensuring trust, coherence, and redistributed authority.

Automated iterative planning for Claude Code. 5-actor verification loop (Researcher → Planner → Skeptic → TDD Reviewer → Critic) that iterates until consensus, producing one-shot-ready implementation plans.

Claude Code plugin — structural guardrails for multi-agent software development. RTM-based evidence, cross-model adversarial audit, 7 MCP tools, worktree isolation, HITL retrospective. DOI: 10.5281/zenodo.19108370

Code for "Loop Closure Detection via Locality Preserving Matching with Global Consensus"

The MATLAB code of "Appearance-based Loop Closure Detection via Bidirectional Manifold Representation Consensus"

Cross-LLM iterative code review plugin for Claude Code. Spawns Codex or Gemini as peer reviewer, loops until consensus.

Dual-agent AI code review loop — consensus at every step, not after the fact. Used to evolve an open-source fork into a production AI assistant with zero manual code.

Official implementation of the IROS'25 paper LoopGNN: Visual Loop Closure Detection Through Deep Graph Consensus

Event-driven microservices auto-scaling system using Kafka and multi-model ML consensus. Services emit metrics, classifiers vote on scaling actions, an authoritative controller aggregates decisions, and Kubernetes executes them. Includes persistence, retraining loop, and fault-tolerant control logic.

Multi-agent debate framework designed to reach high-quality, converged answers through iterative reasoning.

No description available

Multi-model AI consensus loop via OpenCode CLI

Claude Code plugin marketplace — hosts consensus-loop and future plugins

Generates figures for "Quality-controlled R-loop meta-analysis reveals the characteristics of R-loop consensus regions"

Distributed multi-agent system with specialized AI agents, consensus mechanisms, and human-in-the-loop oversight

Cross-model review: Claude Code writes, Codex CLI reviews — automated consensus loop for code plans, novels, specs, and more

Reviewer-first multi-CLI control tower for vibe coding: evidence-gated, memory-aware consensus loops across Codex, Claude, and Gemini.

Multi-agent self-improving loop — BDD-driven, 3-reviewer consensus, dynamic app testing. Claude Code skill for autonomous code quality improvement.

Multi-threaded application that searches for hairpin structures in cDNA / mRNA sequences (FASTA or GenBank) with specific consensus sequences in the loop.

A Novel Expert-in-the-Loop Explainable Engineering Change Proposal Generation Framework: Multi-Agents for Cross-Departmental Information Aggregation and Consensus Reaching

Lightweight OpenAI-compatible LLM gateway — routing, fallback chains, web search/scrape, and multi-model reasoning patterns (consensus, loop, review, devil's advocate, SWOT, brainstorm)

AI-powered SEO article generator with agent-based pipeline — SERP analysis, editorial briefs, hybrid quality scoring, multi-provider consensus (Claude + Gemini), and automated edit loops

Automated iterative planning for Claude Code. 5-actor verification loop (Researcher → Planner → Skeptic → TDD Reviewer → Critic) that iterates until consensus, producing one-shot-ready implementation plans.

Automated eval & improvement loop for AI agents. Generates test scenarios, runs the agent, scores with multi-model consensus (Claude + GPT + Gemini), and iteratively patches code until quality targets are met.

End-to-end platform for managing training data quality at scale: task design, workforce routing, multi-stage annotation, consensus resolution, and model feedback loops. Configurable across AI/ML (RLHF), healthcare, financial services, and e-commerce.

A full-stack FastAPI web application that implements a two-agent reasoning system using Google Gemini models. The system runs a Planner–Critic loop with real-time SSE streaming, multi-turn task refinement, follow-up instructions, auto-run mode, persistent session memory, and consensus synthesis. Includes a custom dark-themed UI built with HTML/CSS/