About this project

Ruflo provides a harness around language models like Claude Code and Codex, giving agents tools, memory, loops, sandboxes, and controls so they can work collaboratively. After a simple `npx ruflo init`, users gain access to over 100 specialized agents organized into swarms, a self‑learning architecture that improves from past tasks, vector‑based memory with HNSW indexing, and a plugin marketplace with more than 30 official extensions covering swarm orchestration, autopilot loops, workflow templates, federation across machines, security auditing, cost tracking, and domain‑specific tools such as neural trading and IoT management. The system includes a CLI, an MCP server for integration with Claude Code, a web UI beta for multi‑model chat with parallel tool calling, and a Goal Planner UI that turns plain‑English goals into executable agent plans using GOAP A* search. Ruflo also supports agent federation with zero‑trust security, enabling agents on different machines or organizations to discover, authenticate, and exchange work while stripping PII and maintaining audit trails. All features are designed to let developers keep writing code while Ruflo handles coordination, memory, and security.