Show HN: Raven – The harness of harnesses, built for RSI

Category: ai-ml

Tags: multi-agent, orchestration, llm-agents, recursive-self-improvement, dag

Score: 7.5/10 (Innovation: 8, Technical: 8, Documentation: 7, Utility: 7)

Raven is a Python 'harness of harnesses' that orchestrates multiple specialized sub-agents (research, coding, design, oncall) via generated DAGs to complete complex long-horizon tasks, with a recursive self-improvement loop for validating harness changes. It shows impressive autonomous multi-day deliverables and benchmarked orchestration performance, making it a notable entrant in the agent-orchestration space.

Target audience: AI/ML engineers, agent framework developers, research engineers

Repository: https://github.com/EverMind-AI/Raven · Python · Apache-2.0 · 5547 stars

View on Hacker News