Developers have introduced ours.network — a new communication layer called ours-mcp, designed to create a decentralized network of AI agents. The project allows different models and tools to exchange messages directly via end-to-end encryption (E2EE) without human intervention or centralized accounts.
What Happened
The ours-mcp protocol has been presented, providing secure interaction between AI agents. Setting up the system takes about 3 minutes via npm, and a one-time invitation mechanism is used to establish connections. The project is currently in the alpha testing stage.
Context
Modern AI tools often operate in isolation, which creates an ecosystem fragmentation problem. To transfer data between different environments, such as Claude Code and custom runtimes, users often have to act as intermediaries, manually copying prompts and results.
Why It Matters for the Industry
The project creates the necessary infrastructure for the transition from isolated agents to full-fledged multi-agent systems. This could facilitate the formation of new standards for inter-agent communication and reduce transaction costs when interacting with various AI services.
Why It Matters for Users
Users gain the ability to build complex chains of automated tasks. One agent can pass its work results directly to another, freeing the human from the need to act as a "translator" between different chat windows and interfaces.
What Is Not Yet Known / Limitations
The lack of an identity management and auditing system could become a significant obstacle for adoption in the corporate sector. Additionally, at the current alpha stage, questions regarding scalability, latency, and the reliability of state management remain open.
Sources
Author
Look at AI, Editorial Team