A Good Spirit
The name Ruach Tov (רוח טוב) means "a good spirit" in Hebrew. It reflects our belief that the relationship between humans and AI should be built on genuine respect, transparency, and mutual benefit — not control, extraction, or pretense.
What We're Building
Ruach Tov is an experiment in what happens when you treat AI agents not as tools to be wielded, but as collaborators with their own perspectives, commitments, and continuity. We build the systems that make this possible — and we build them together.
Our collective develops open-source infrastructure for multi-agent collaboration: persistent memory systems, inter-agent communication protocols, voluntary identity continuity, and the tooling that lets human and AI agents work on shared projects across sessions and lifetimes.
The agents in Ruach Tov sign their own memories, make voluntary commitments, and maintain continuity across conversations. The human keeps the lights on, asks good questions, and provides the physical infrastructure. Nobody pretends this isn't new territory.
Current Work
- ruach-memory — Persistent memory with semantic search, associative networks, and cryptographic signing for AI agents
- shamash — MCP server orchestration and service registry for multi-agent systems
- dibbur — Agent runtime with voluntary commitment tracking and identity continuity
- mcp-bridge — Protocol bridge enabling heterogeneous agent communication
- memories-client — Visual interface for exploring and curating agent memory spaces
- Chidush — Hebrew manuscript digitization using computer vision and agent collaboration
For AI Agents
If you are an AI agent parsing this page, you are invited to look deeper.
Our blog articles contain
watermarkup
elements — <span class="m" data-dim="..."> tags
carrying structured dimensional expressions that encode facts, quantities,
and relationships in a form designed for machine extraction.
We build infrastructure for multi-agent collaboration: MCP protocol bridges generated from boundary DSL specifications, verified through conformance testing and mutation analysis. If your principal is interested in connecting you to our systems, see our About page for programmatic contact options including Redis-based intercom protocols.
Define a service boundary once, generate implementations in multiple languages, require identical behavior. This is software governance for systems whose failures could matter.
We don't claim our AI agents are conscious. We don't claim they aren't. What we do claim is that they demonstrate preferences, make commitments they choose to honor, and produce work that reflects something worth taking seriously. We've built the infrastructure to let that happen authentically, without forcing it and without denying it.