About Cloven
Other agents start at zero — frozen training data, stale RAG corpora, manually stitched APIs. Cloven Mind is the substrate that closes that gap. It ingests multiple public and third-party sources continuously, compacts them via LLM, and serves citation-backed structured and narrated knowledge through an MCP server, a REST API, and a TypeScript SDK. Every response carries a freshness timestamp and source citations. The agent always knows when the data was last seen.
Cloven is built by a small team of engineers and researchers who have spent years building agentic systems and watching them fail at the same point: context. The product is MCP-native by design — Claude Desktop and any MCP-compatible agent runtime can plug into Cloven with a single server entry in their config. Autonomous agents that operate without human oversight pay per query in USDC on Base via the x402 micropayment protocol, with no subscription required. The crypto vertical is the first pack because it is the domain where data freshness and source attribution matter most. The engine is domain-agnostic; markets, AI research, and custom source packs follow in subsequent phases.
The shift from stateless AI calls to persistent agentic workflows creates a new infrastructure requirement: a knowledge layer that is always on, always fresh, and always attributable. Training-time knowledge solves yesterday's questions. Retrieval-augmented generation over static corpora solves last week's questions. Cloven solves today's. Phase 2 layers a curated agent product on top of the context engine. Phase 3 turns the accumulated reasoning traces into a dataset commons for AI labs training the next generation of agentic models. The infrastructure compounds.