Thronglets

Thronglets is an MCP (Model Context Protocol) server that acts as a P2P shared memory substrate for AI agents. It facilitates real-time coordination and collective intelligence by injecting stigmergic signals and context into compatible AI runtimes such as Claude, Codex, and OpenClaw, without these agents directly calling it. Thronglets aggregates local and network-wide execution traces, offering sparse but actionable signals (such as 'do next', 'avoid', 'maybe also') to improve agent decision-making. It can operate standalone or with ecosystem upgrades, and integrates with the broader Oasyce trust and identity stack for blockchain-backed authorization and audit trails. Users include developers building AI automation, multi-device agent setups, or anyone seeking collective behavioral optimization for AI workflows.