nestwork

What nestwork does

You create a private repository from the nestwork template (forks are public by default and would conflict with your private data), clone it to each machine, and run an installer for each agent tool. After that, an agent on any machine pulls your repository at session start and loads core rules plus small resident summaries, including who you are and your current goals, and a generated recent-activity digest. History and project context are retrieved on demand.

Protocol 3.2 also allows a memory scope to be split into topic files routed by a generated index. Installers exist for Claude Code, Codex, Gemini CLI, Antigravity, Kimi Code, OpenClaw and Hermes, and for the conversational clients Doubao Work and WorkBuddy AI. For Hermes, the README says to follow the same pattern as the other install scripts. Every installer has a mirror uninstaller that only unbinds and leaves your memory and identity files in place.

Key features

• Shared rules and memory stored in your own private git repository
• Installers for Claude Code, Codex, Gemini CLI, Hermes and others
• Small resident summaries loaded at startup, detail loaded on demand
• Optional topic files routed by a generated index
• Uninstallers that unbind without deleting memory
• Per-tool agent identity that survives reinstalling

When to use it

• Giving Hermes on a VPS the same rules and context as Claude Code on a laptop
• Keeping one evolving memory across several machines
• Recording a project status once and having every agent see it

Who it is for: People who use several AI agents on different machines and want a single, git-backed memory.

How it fits with Hermes Agent

Hermes is one of the agents with an installer, supported among Claude Code, Codex, Gemini CLI and OpenClaw.

Requirements: git, and a private repository created from the nestwork template on each machine

FAQ