Moss

Moss develops a real-time semantic search runtime aimed at conversational and multimodal AI applications. Rather than operating as a hosted vector database, the product embeds retrieval and embedding generation inside the customer's own process, removing the network round trip that the company says typically adds 200-500 ms to a query. Moss states end-to-end retrieval latency of under 10 ms and positions the runtime for voice agents, copilots, chat interfaces, documentation and knowledge search, and offline or on-device applications where retrieval sits on the critical path.

The system supports hybrid retrieval combining semantic and keyword search, built-in embedding models (with the option to supply an external model), metadata filtering with operators such as $eq, $and, $in and $near, and a managed cloud layer (Moss Cloud) for storing, syncing and distributing indexes. SDKs are published for Python (3.10+), TypeScript/Node.js (20+), Elixir and C (libmoss), alongside a separate WebAssembly package (@moss-dev/moss-web) for client-side search in the browser, a CLI for index management, and data connectors for SQLite, MongoDB, MySQL and Supabase. Y Combinator's profile describes the underlying vector index as built in Rust and WebAssembly and running natively across browsers, mobile devices and servers.

Moss publishes framework integrations for LangChain, DSPy, LlamaIndex, Pipecat, LiveKit, Vapi, ElevenLabs, Strands Agents, CrewAI, Haystack, AutoGen, Mastra, Langflow, Pydantic AI and the Vercel AI SDK, and its open-source repository includes worked examples for voice agents (an airline PNR ambient-retrieval demo and a multi-agent mortgage lending flow), classification, custom embeddings and browser/WASM use.

Founding story

Founded in 2024 by Sri Raghu Malireddi (Founder and CEO) and Harsha Nalluru (co-founder and CTO), who had known each other for more than eight years. Sri previously led ML work at Grammarly and Microsoft, shipping LLM and personalization systems across Office, Bing and Grammarly, including personalization work cited as driving 300% retention growth for Grammarly Keyboard and scaling models to 40M+ DAUs; he has published at conferences including ACL and holds patents in real-time ML. Harsha was a tech lead at Microsoft where he architected the core Azure SDK stack supporting 400+ cloud services and 100M+ weekly npm downloads. The founders say the idea came from repeatedly encountering retrieval lag while building large-scale agentic systems at Microsoft and Grammarly [7].

Business model

Moss distributes SDKs and an open-source repository while operating a managed cloud service; developers sign up at moss.dev for a project_id and project_key, with a free tier available and self-serve onboarding described as requiring no credit card. Indexes can be created and managed through a web portal or programmatically via the SDKs and CLI [0][2][7].

Sources describe a free tier plus paid usage tied to project credentials for the managed Moss Cloud index layer; specific pricing is not disclosed in the available material. The company reported three paying customers at the time of its Y Combinator launch [2][7].

Traction

The company website reports 250,000+ installs and use in production by teams building real-time AI systems. The public GitHub repository shows 666 stars, 91 forks and 257 commits. At its Y Combinator launch, Moss reported six enterprise design partners, three paying customers and seven further prospects evaluating, with usage and revenue described as growing roughly 100% week over week, plus production pilots achieving sub-10 ms retrieval and 70-90% token savings versus traditional pipelines [0][2][7].

Latest developments

Moss publishes a reference architecture guide, "The Production AI Stack," and a live latency demo environment on its site. The GitHub repository documents newer capabilities including Elixir and C SDKs, a browser/WASM SDK, database connectors for SQLite, MongoDB, MySQL and Supabase, a CLI, and voice-agent example applications. The company is hiring a Founding Customer Success Manager and an SDK Software Engineer [0][2][7].