The original open-source agent meta-harness, behind 12.5 million npm downloads, now turns one mission into an end-to-end agent workflow and lets teams orchestrate swarms, intelligence, safety and operations from one place, with human approval where it matters.

SAN FRANCISCO, CA / ACCESS Newswire / October 13, 2026 / A single AI coding assistant can answer a question or make a change. Real engineering work needs more: research, code, tests, security review, and a record of why decisions were made. Cognitum One™, Inc. (cognitum.one) today announced the latest release of Ruflo™, the open-source AI agent meta-harness and multi-agent orchestration platform for Claude Code, Codex, and other Model Context Protocol (MCP) clients. Ruflo turns one assistant into a coordinated team of more than 100 specialized agents that plan, build, test, and check their work in parallel. Ruflo from Cognitum One adds two core capabilities: Ruflo Mods, which turn one mission into an end-to-end agent workflow that researches, builds, tests, validates, secures, benchmarks, and learns; and a single control plane to orchestrate swarms, intelligence, safety, operations, and extensible networks, with human approval where it matters.

Ruflo is built on a simple premise: an agent is a model plus a harness. The model reasons and writes; the harness supplies the tools, memory, workflows, sandboxes, and controls that let it do real work. Ruflo is that harness, and the original one built for the job. It organizes agents into queen-led swarms, gives them persistent shared memory, routes each task to the right model, and records what ran, what it cost, and whether it passed. The project has earned more than 74,000 GitHub stars and 8,800 forks, and the Ruflo ecosystem recorded 12.5 million npm downloads in the twelve months. Ruflo also powers agent coordination inside ruOS, Cognitum One’s enterprise computer platform for AI agents.

“Most people meet AI through a chat window, but real work rarely ends with one answer,” said Reuven “rUv” Cohen, Founder and Chief AI Officer of Cognitum One and creator of Ruflo. “With Mods, you hand Ruflo a mission and it runs the whole loop: research, build, test, validate, secure, benchmark, and learn from the result. The console shows you every step, and you decide where Claude acts on its own and where it has to ask. Federation then extends that team across machines, with outbound data gated by the policies you set.”

What’s New in Ruflo

  • Ruflo Mods: one mission, end to end. Mods turn a single mission into a complete agent workflow that researches, builds, tests, validates, secures, benchmarks, and learns. Mods route prompts and enforce policy directly inside Claude Code.
  • One control plane, with human approval. Teams orchestrate swarms, intelligence, safety, operations, and extensible networks from one place. Ruflo Console, a visual cockpit beside Claude Code, keeps missions, workflows, agent activity, logs, token usage, and cost in view. Administrators choose read, write, manage, or full control for Claude and whether it acts automatically or asks first; the default is read access with approval, and Take back control pauses Claude’s console tools at any time. An optional Orbital theme adds a rotating globe, swarm routing previews, and six palettes, with the classic look and reduced motion available.
  • Cross-machine agent federation. Independent Ruflo installations on different computers, organizations, or cloud regions can discover each other, delegate sub-tasks, and share workloads. Remote agents start untrusted and prove identity through mTLS and Ed25519 challenge-response, without shared secrets. Behavioral trust scoring upgrades reliable peers over time and downgrades misbehaving ones immediately.
  • Built-in PII data gating. A configurable 14-type detection pipeline scans outbound federation messages for data such as email addresses, API keys, and Social Security numbers before they leave the local machine. Administrators set per-trust-level policies to block, redact, hash, or pass each type, and HIPAA, SOC 2, and GDPR modes produce structured, searchable audit records.
  • SONA self-learning memory. Agents improve without retraining the underlying model. SONA, working with ReasoningBank and trajectory learning, captures successful execution patterns as local embeddings in HNSW-indexed AgentDB memory, so agents can reuse what worked in later tasks.
  • 314 native MCP tools. Ruflo’s Model Context Protocol surface spans 314 tools covering continuous integration checks, local security scans, cloud deployment hooks, and observability pipelines, plus 26 CLI commands and 35 plugins.
  • Adaptive backpressure. A new messaging layer regulates delivery speed between swarm nodes, preventing timeouts and agent crashes during long, high-volume software engineering jobs.
  • MetaHarness and ruflo eject. MetaHarness grades an agent setup’s readiness from 1 to 100, scans tool configurations for security risks, and snapshots projects to catch regressions between runs. ruflo eject packages a configured swarm into a standalone agent toolkit with its own name and command-line package.

Core Capabilities

  • Swarm coordination. Hierarchical, mesh, and adaptive topologies with Raft, Byzantine, and Gossip consensus, with queens assigning work to architect, coder, tester, reviewer, security, and documentation agents.
  • Intelligent model routing. Simple tasks go to faster, lower-cost models and complex reasoning to top-tier models, across Claude, GPT, Gemini, Cohere, and local Ollama models with automatic failover.
  • Structured methods. SPARC (Specification, Pseudocode, Architecture, Refinement, Completion), architecture decision records, and domain-driven design keep multi-agent work reviewable.
  • Security controls. AIDefence blocks prompt injection and detects PII, alongside input validation, CVE remediation, and cryptographic verification of installed packages with ruflo verify.

Built for Enterprise Teams

Ruflo works alongside the tools and infrastructure enterprises already run. Two business units could use federation to share fraud signals while the PII pipeline strips customer data from every outbound message. A platform team could build a governed swarm for its developers, then use ruflo eject to ship it under its own name. Organizations that need managed, centrally governed computers for their agents can run Ruflo inside ruOS™ on infrastructure they control.

“Every enterprise we talk to is asking the same question: how do we move from one helpful assistant to a team of agents we can govern, measure, and trust?” said D. Christopher Keil, Founding CEO of Cognitum One. “Ruflo answers it. A team hands over a mission and gets back an end-to-end workflow, with approvals where their policies require them, a record of what ran and what it cost, and data controls that hold across teams and machines. Customers can start with a single command and grow into federated deployments on their own infrastructure, alongside ruOS.”

Getting Started

  • Claude Code: run /plugin marketplace add ruvnet/ruflo, install ruflo-console@ruflo and ruflo-mods@ruflo, run /reload-plugins, then open the console with /ruflo.
  • Codex and other MCP clients: run npx ruflo@latest init wizard in a project, then connect the client to the Ruflo MCP server with npx -y ruflo@latest mcp start.
  • Federation: install ruflo-federation@ruflo, then use federation init and federation join to connect installations.

Requirements and current installation details are in the Ruflo guide on GitHub. Cognitum One recommends starting with one bounded task, such as a read-only project review, and adding agents, memory, and federation as each one earns its place.

Availability

Ruflo is available today. Ruflo is open source under the MIT license at github.com/ruvnet/ruflo; model provider subscription or usage costs still apply. Ruflo from Cognitum One gives organizations end-to-end mission workflows through Ruflo Mods and one control plane for swarms, intelligence, safety, operations, and extensible networks, along with commercial licensing, enterprise support, and deployment with ruOS. To learn more, visit cognitum.one or download the product at https://www.producthunt.com/products/ruflo-console.

Resources

About Cognitum One

Cognitum One is an artificial intelligence company building a sovereign agentic platform that orchestrates AI across the entire compute plane, from the public cloud to on-premises systems to the low-power edge, placing every workload where performance, cost, and security require. The platform is private by default, with real-time self-learning memory, distributed swarm coordination, contactless perception, cryptographic attestation, and native MCP integration, plus SDKs for Rust, Node.js, and Python. It is built for enterprises operating under data sovereignty and regulatory constraints. The company is led by Founder and Chief AI Officer Reuven Cohen, creator of the widely used open-source AI projects RuView™, RuFlo, and RuVector™, and Founding CEO D. Christopher Keil. To learn more, visit cognitum.one or explore the platform repositories on GitHub.

Claude and Claude Code are trademarks of Anthropic, PBC. OpenAI and Codex are trademarks of OpenAI. Cognitum One is not affiliated with or endorsed by Anthropic or OpenAI. Star, fork, agent, tool, and plugin figures are from the public ruvnet/ruflo GitHub repository as of October 12, 2026; the download figure covers Ruflo ecosystem npm packages over the twelve months through October 4, 2026.

Media Contact
Nick Ruest
Chief of Staff, Cognitum One, Inc.
nick@cognitum.one
(416) 477-3336

# # #

SOURCE: Cognitum One

Related Documents:

View the original press release on ACCESS Newswire

Media gallery

About The Author
Cognitum One