ruflo: Coordinate AI Agents and Autonomous Workflows

Overview

Ruflo is an agent harness that adds an execution and coordination layer around Claude Code and Codex. It is intended to help agents collaborate on tasks, retain useful context across sessions, and run structured workflows rather than operate as isolated assistants.

The project combines a CLI setup path with a plugin marketplace. It may suit developers and teams that want agent orchestration, memory, and extensibility in their coding workflow. It is a substantial system, so users seeking only a lightweight single-agent helper may find its breadth unnecessary.

Key Features

• Coordinates multiple agents using hierarchical, mesh, or adaptive swarm topologies.
• Provides specialized agent definitions for tasks such as coding, testing, security, documentation, and architecture.
• Adds persistent vector memory and learning components for retrieving and reusing task patterns.
• Exposes integrations through an MCP server and Claude Code plugins.
• Supports reusable workflows, hooks, and background workers for automating recurring work.
• Includes federation features for agent communication across installations, with identity and data-handling controls described in the project documentation.
• Offers provider integrations including Claude, GPT, Gemini, Cohere, and Ollama.

Use Cases

• Coding teams: Coordinate implementation, testing, and review agents on larger changes when manual handoffs become cumbersome.
• Claude Code or Codex users: Add reusable agent roles, memory, and workflow automation to an existing coding assistant setup.
• Workflow builders: Package repeated multi-step development processes as workflows and invoke them through CLI or MCP integrations.
• Distributed agent projects: Explore cross-installation collaboration when agents need to exchange work under explicit trust and security controls.

Getting Started

For the full CLI setup, run:

npx ruflo@latest init wizard

See the README for plugin installation, platform-specific guidance, configuration, and usage details.

Alternatives

• oh-my-claudecode: oh-my-claudecode focuses on orchestrating agents within Claude Code, while Ruflo also targets Codex and adds persistent memory and MCP integrations.
• clowder-ai: Clowder coordinates agent CLIs across model families with shared roles and memory, while Ruflo is a TypeScript harness centered on Claude Code and Codex.
• CLI Agent Orchestrator: CLI Agent Orchestrator delegates among coding CLIs through a local server and tmux sessions, rather than providing Ruflo’s TypeScript harness and persistent memory.
• Orkas: Orkas is a local-first desktop app that runs coding CLIs with a Commander agent, while Ruflo is a TypeScript harness with workflow automation and MCP integrations.

Considerations

• The project offers two distinct paths: plugins for a lighter setup and CLI initialization for the full harness. Their capabilities and installed files differ, so choose based on the required integration depth.
• The README describes a broad feature set and many plugins. Review the relevant documentation and test the specific capabilities your workflow depends on.
• The README warns that Claude Code mods run with the permissions of the user's account and are not sandboxed. Read their code before installing them.
• Federation and security claims describe project features, not independent verification. Evaluate data handling and trust requirements before using them with sensitive workloads.
• Ruflo targets Claude Code and Codex workflows. Teams using other environments should verify integration support before adopting it.