Moss is a search runtime designed for voice agents, copilots, and multimodal applications, delivering sub-10ms lookups with no infrastructure management. Built using Rust and WebAssembly, it provides real-time retrieval capabilities for apps, browsers, and enterprise agents. The process is streamlined into three simple steps: first, push documents or chat histories to Moss through its SDK or portal; second, Moss creates a compact index and deploys it where your agent operates—be it on a browser, edge device, or cloud; third, the agent retrieves context directly, ensuring minimal latency. This tool targets developers creating AI experiences where speed is critical. Moss allows for local data handling, enhancing privacy and performance. It’s lightweight, with an engine under 20kB, making it suitable for both mobile and desktop applications. The platform also supports features like analytics management and optional rollouts, which can be accessed via a cloud dashboard, ensuring teams maintain visibility without the hassle of infrastructure maintenance. Moss is already in use by over 500 teams, demonstrating its effectiveness in various scenarios. It enables instant recall of user context, fast search capabilities within help centers, and smart searching in note-taking apps—all without sending data online. The continuous improvement of search quality, along with built-in A/B testing for tuning indexes, allows teams to optimize their retrieval results effectively. Each user's data is embedded and updated locally, making interactions feel faster and more personalized while keeping everything secure.