NCP (Natural Context Provider) is a unified MCP that centralizes tool access for AI assistants. Instead of juggling 50+ disparate MCPs, your AI sees just two simple tools: find for intent-driven discovery and run to execute any discovered tool. NCP manages all underlying MCPs by routing requests, discovering the best match, caching responses, and monitoring health. It enables on-demand loading of tools, reducing context size and improving response times. It supports project-level configuration and automatic per-project MCP detection, allowing teams to define local MCPs in each project while falling back to a global configuration when needed. It offers multiple transport options, including stdio for local MCPs and HTTP/SSE for remote services, and supports hibernation-enabled servers to minimize idle costs. NCP also includes smart health monitoring to detect broken MCPs and automatically bypass them, ensuring the AI keeps progressing. The system emphasizes token efficiency, faster tool selection, and a consistent interface across clients (Claude Desktop, Claude Code, Cursor, VS Code, etc.).