Tag
NavMCP is an agentic scaffolding framework that integrates vision-language models with navigation foundation models to enable persistent long-horizon physical-world exploration, achieving state-of-the-art results on benchmarks and real-robot tasks.