Early-stage software. Shurli is experimental and built with AI assistance. It will have bugs. Not recommended for production or safety-critical use. Read the disclaimer.

About

What is Shurli?

Shurli is AI-native peer-to-peer networking infrastructure. It connects your devices directly through NAT, CGNAT, firewalls, and across networks, without relying on cloud services or vendor accounts.

Today it connects your devices. Tomorrow it connects your agents. The same encrypted P2P layer that lets you reach your home server is the foundation for autonomous agent-to-agent communication, with no cloud middleman and no central authority.

The Zero-Human Network

Zero-human companies are coming. They need a network that operates itself, where agents connect, negotiate, and transact directly. Intelligence at every node, not at a central server.

Shurli is being built as that network. Read more about the vision.

Engineering Philosophy

Shurli follows a few core principles. For the full story, including how these principles evolved and their intellectual lineage, see the Development Philosophy page and the blog post on how we build.

Self-Sovereignty First
Your keys, your peers, your network. No accounts to create, no services to subscribe to, no data leaving your control. The trust model is SSH-style: an authorized_keys file that you manage directly.
Single Binary, Zero Dependencies
One file to install, one file to run. No Docker, no runtime, no package manager needed. This isn’t just convenience - it’s resilience. Fewer moving parts means fewer things that break.
Honest About Limitations
When Shurli can’t establish a direct connection, it tells you. When it falls back to relay, it tells you. No silent degradation, no hidden costs.
Docs as First-Class Deliverable
Every architecture decision is documented with the reasoning behind it. The Engineering Journal captures not just what was built, but why every choice was made.

AI-Native Development

Shurli is not a project that bolted on AI tooling. It is built AI-native from day one: architecture, code, documentation, and testing, all developed with AI as a core part of the process.

The direction and decisions are human. The execution leverages AI at every layer. Every line is reviewed, tested, and shipped with the same rigor regardless of origin. The code speaks for itself.

Open Source

Shurli is open source under the MIT license. Contributions, questions, and feedback are welcome. File an issue or open a pull request.