rUv is the public identity of AI engineer Reuven Cohen, founder and chief AI officer of Cognitum One. His open-source work covers coordinating agents, retaining their context and deploying systems on controlled infrastructure.[1]
Start with Ruflo →
Links
GitHub Personal site Cognitum One 
X Founder and Chief AI Officer · Cognitum One
Career
Cohen’s GitHub biography connects the ruvnet account to Cognitum One and the ruv.io site. It describes the public repositories as an ecosystem containing experiments, reusable libraries and community contributions, while Cognitum One provides commercial deployment and governance. That distinction is useful when reading the work: repository ownership identifies the publishing account, but does not make every commit or imported component an individual invention. The current public remit spans agent orchestration, memory and execution across cloud and edge environments.[1]
Contributions
An execution layer around a model
Ruflo, formerly Claude Flow, supplies coordination around coding agents. A model produces text or a proposed action; the harness gives that model tools, task state and a loop in which to work. Ruflo adds facilities for coordinating multiple agents, shared context and a console for observing activity. Its README distinguishes the Claude Code-specific console and runtime modifications from the MCP server used by other clients, so those installation paths should not be treated as interchangeable.[2]
Keeping useful context between sessions
RuVector supplies persistent vector retrieval and graph relationships for agent memory. Vector search finds stored material whose representation is similar to a query, while graph links can preserve relationships between pieces of context. The repository combines those retrieval mechanisms with explicit feedback and memory lifecycle controls. Its examples show storing and recalling a decision locally. This gives an agent a way to recover prior project information instead of expecting every session to begin with the same complete prompt.[3]
Tools, writing & talks
Tools
Start here
Ruflogithub.comAgent coordination and execution tooling around Claude Code and Codex, with client-specific interfaces and an MCP server.
Begin with Ruflo’s explanation of model versus harness, then choose the setup path for the client you actually use. Command-line experience and familiarity with coding agents help; inspect optional integrations before enabling them.
Project lead and repository owner[1][2]Original source- GenAI Superstreamgithub.com
Agentic engineering toolkit in Python for building AI-powered data analysis workflows.
Catalog workOriginal source - Marketing Swarmgithub.com
Agentic swarm framework for running AI-driven marketing campaigns and content workflows.
Catalog workOriginal source - SAFLAgithub.com
Python framework for building self-improving AI agents around a self-aware feedback loop with reflective memory.
Catalog workOriginal source - QuDAGgithub.com
Quantum-resistant protocol for anonymous communication built on a DAG, implemented in Rust.
Catalog workOriginal source - SPARCgithub.com
Five-stage prompting method for AI-assisted development, running from specification through to completion.
Catalog workOriginal source - drupaljsgithub.com
JavaScript port of Drupal patterns bringing node-based content modeling and rendering to the Node.js ecosystem.
Catalog workOriginal source - Code Meshgithub.com
Multi-agent orchestration system where coding agents collaborate over a shared mesh of edits and tests.
Catalog workOriginal source - Agentic Securitygithub.com
Framework for security-focused AI agents that automate reconnaissance, vulnerability triage, and exploit explanation.
Catalog workOriginal source - dynamo-mcpgithub.com
Dynamic MCP registry that spins up custom MCP servers from parameterized Cookiecutter scaffolds on demand.
Catalog workOriginal source - ARCADIAgithub.com
AI-powered game engine for dynamic, personalized experiences in evolving worlds, written in Rust and Node.js.
Catalog workOriginal source - agentdbcognitum.one
Vector memory for agents with feedback-aware retrieval that learns which results actually helped.
Catalog workOriginal source - Reflective Engineergithub.com
Coding agent that reasons about its own reasoning before each move and commits to actions with explicit justification.
Catalog workOriginal source - Federated MCPgithub.com
Implements MCP federation against the official spec, covering message framing, transport, and the full lifecycle.
Catalog workOriginal source - Agent Name Servicegithub.com
Secure discovery and interaction protocol for AI agents across organizations, part of the OWASP GenAI Security Project.
Catalog workOriginal source - Sublinear-Time Solvergithub.com
Rust and WASM solver for asymmetric diagonally dominant linear systems with an npm CLI and HTTP API.
Catalog workOriginal source - Ultrasonicgithub.com
Hides agent-to-agent commands inside audio and video at ultrasonic frequencies, then pulls them back out.
Catalog workOriginal source - symbolic-scribegithub.com
Note-taking tool that captures structured knowledge with named relations between concepts for graph-based recall.
Catalog workOriginal source - ruDevolutiongithub.com
Rust decompiler that recovers original names with AI and proves each transformation with cryptographic witness chains.
Catalog workOriginal source - auto-browsergithub.com
Command-line web automation tool that drives sites from YAML configs with AI-assisted template creation.
Catalog workOriginal source - Flow Nexusflow-nexus.ruv.io
Agentic platform built on MCP where users deploy AI swarms, compete in coding challenges, and earn rUv tokens.
Catalog workOriginal source - RVMgithub.com
Virtual machine written in Rust that runs AI-generated code in a deterministic, isolated runtime.
Catalog workOriginal source - voicebotgithub.com
Speech-to-speech agent handling real-time spoken conversation with LLMs, implemented in Python.
Catalog workOriginal source - MidStreamruv.io
Rust platform that interprets what an AI is saying mid-stream instead of waiting for the full response.
Catalog workOriginal source - FACTgithub.com
Retrieval pattern that skips vector search by caching static tokens in the model and fetching live facts on demand.
Catalog workOriginal source Framework for building self-managing applications from decentralized agents that coordinate without a central planner.
Catalog workOriginal source- dspy.tsgithub.com
TypeScript port of DSPy for optimizing LLM pipelines programmatically from examples instead of prompt tweaking.
Catalog workOriginal source - SynthLanggithub.com
Prompt language that compresses LLM prompts about tenfold using logographic scripts and symbolic constructs.
Catalog workOriginal source - open-claude-codegithub.com
Nightly decompiled and reverse-engineered rebuild of the Claude Code CLI as an open-source project.
Catalog workOriginal source - ruv-FANNgithub.com
Memory-safe neural network library for Rust that modernizes the classic Fast Artificial Neural Network library.
Catalog workOriginal source - RuVectorgithub.com
A Rust-based substrate for vector retrieval, graph relationships and persistent agent memory.
Creator credited by repository[3]Original source - RuViewgithub.com
Estimates human pose, breathing, and presence from commodity WiFi signals, with no cameras or wearables.
Catalog workOriginal source
Writing, research & talks
- rUv-devgithub.com
rUv's AI-powered development methodology — the rUv approach: prompts, patterns, and skills for shipping with AI agents.
Catalog workOriginal source