What is foundation-protocol?

Foundation Protocol is a Python library for building systems where AI agents, human users, and software tools can all communicate with each other through a shared set of rules. The underlying idea is that today these three types of participants often need custom glue code to work together, and this library provides a common layer so that any entity, whether it is a person, a running AI model, or a third-party tool, can be addressed and messaged in the same way. The library gives each participant a registered identity on a host node. You create a host, register entities on it with their type (agent, human, tool, or service), and then those entities can send structured messages to each other using an async Python interface. The library handles the session lifecycle, meaning the rules about when a collaboration starts, how events are recorded so they can be replayed, and when a session ends. Beyond basic messaging, the library includes a governance layer that lets you define checkpoint hooks: points in a workflow where you can verify who did what and whether they were authorized to do it, creating an audit trail. There is also a built-in system for financial transactions between entities, covering contracts, holding funds in escrow, settling payments, and resolving disputes through state machines. Finally, the library supports federation, meaning entities registered on different host nodes on different machines can route messages to each other across the network. The project accompanies an academic paper published on arXiv and has full documentation online covering quickstart guides, the checkpoint pipeline, trade and trust mechanics, and security boundaries. A full-stack application called AI-Link-Net, listed in the README, is built on top of this protocol and demonstrates it in a complete system. Installation is via pip directly from the GitHub repository, and example code in the repository covers cross-host messaging, tool integration, and payment workflows.