Ruflo

Ruflo is an open-source agent harness with 77,000+ stars on GitHub — the original framework for deploying multi-agent swarms that coordinate autonomous workflows.

Overview

Ruflo began as a pioneering experiment in agent orchestration and grew into a full harness for multi-player agent systems: instead of one agent working alone, you deploy swarms of specialized agents that divide tasks, coordinate through shared memory and check each other’s work. It features adaptive memory, self-learning behavior, federation across agents, vector RAG integration and an MCP server interface, and it plugs natively into Claude Code, Codex, Hermes and other agent environments. The TypeScript-based project is MIT licensed and intensely active, with repository updates in October 2026. For builders who think in teams of agents rather than single chatbots — automated development crews, research swarms, conversational AI systems — Ruflo provides the coordination layer, the memory and the workflow patterns to run them.

Key Features

• Multi-agent swarms – deploy coordinated teams of specialized autonomous agents.
• Adaptive memory – shared, self-learning memory keeps swarms consistent over long tasks.
• Vector RAG integration – agents ground their work in your documents and data.
• Federation – agents can coordinate across instances and environments.
• Native integrations – works with Claude Code, Codex, Hermes and MCP-based tooling.
• Conversational AI systems – patterns for building multi-party conversational products.

💳 Pricing

Ruflo is free and open source under the MIT license. You pay only for the model API usage of the agents you run through it.

Who Is It For

Developers building multi-agent systems — automated coding crews, research swarms and coordinated autonomous workflows — who need orchestration, memory and federation rather than another single-agent chat loop.