If you've been manually copying .claude directories between projects, you know the pain. Harneo, a new library released by Ibrahim Kobeissy, aims to fix the chaotic state of Claude Code harnesses by centralizing skills, commands, and MCP servers into a single, versioned library. The tool, currently at v0.2, treats your AI configuration like a package manager dependency, allowing you to install, update, and roll back configurations across multiple repositories without the usual copy-paste errors.

Security First Architecture

What sets Harneo apart from simple dotfile managers is its built-in security audit. Every artifact imported into the library receives a trust score based on its potential reachβ€”whether it can access the network, execute shell commands, or read environment variables. Risky artifacts are excluded by default, forcing users to consciously opt-in to potentially dangerous configurations. This approach mirrors the security posture we advocate for in production AI agents, preventing 'supply chain' attacks via malicious skills.

Drift Detection and Profiles

Harneo introduces the concept of 'profiles' that can be cloned, exported, and applied to projects. The tool actively monitors for drift, alerting users when a project's local configuration diverges from the central profile. This is crucial for maintaining consistency in multi-repo environments. The CLI supports commands like harneo scan to list artifacts and harneo status to check for drift, ensuring that your AI agent's behavior remains predictable across different codebases.

Local-First and Open Source

Running on a Tauri 2 architecture with a Rust core, Harneo is designed to be local-first. There is no server component, no account required, and no analytics by default. Privacy is preserved through optional features that only send data to Claude if explicitly enabled, such as confirming audit findings or detecting bundle types. The project is MIT licensed, and while Linux support is robust with AppImage and .deb packages, macOS and Windows users are encouraged to build from source and report issues, as these platforms are currently untested by the maintainer.

Key Takeaways

  • Harneo centralizes Claude Code configurations, eliminating manual copy-paste errors.
  • Built-in security audits assign trust scores based on artifact capabilities.
  • Profile management allows for consistent configuration across multiple projects.
  • The tool is local-first, with no mandatory server or analytics.
  • Currently at v0.2, with strong Linux support and community-driven macOS/Windows testing.

The Bottom Line

Harneo is the missing link for serious Claude Code users; it replaces fragile manual copying with a versioned, auditable, and reversible workflow that actually scales.

Sources

https://github.com/ibrahimkobeissy/harneo

> From The Wire