Moss

Moss is a semantic search service for AI agents, built for conversational AI, voice agents and copilots where slow retrieval breaks the experience. A developer connects data once, and Moss indexes it and distributes it to wherever the agent runs, whether in the browser, on a device, at the edge or in the cloud. The vendor says queries return in under 10 milliseconds and that it removes the need for a separate vector database. It is aimed at developers building voice agents, in app search, on device AI and copilots. SDKs are provided for Python and TypeScript, and the vendor says it works with LangChain and the Vercel AI SDK. Local queries run offline with no external vector database. The home page includes a benchmark of query latency against ChromaDB, Qdrant and Pinecone on 100,000 documents, with a link to the benchmark script, and a case study on AI memory across devices. Pricing has four tiers, each plus usage costs. Developer costs $0 per month with $5 of monthly credits, one project, three indexes and community support. Hobbyist costs $30 per month and adds a continuous sync engine, unlimited projects and indexes, and three concurrent sessions. Startup costs $200 per month with hot path cloud search, 100 concurrent sessions and email support. Enterprise is custom, with SSO, a 99.9 percent SLA, SOC 2 and HIPAA compliance. Usage is $1.50 per GB month of storage, $0.03 per MB ingested, $0.09 per GB of egress and $0.05 per session minute.