// the find
alibaba/nacos
an easy-to-use dynamic service discovery, configuration and service management platform for building AI cloud native applications.
Nacos is a service registry and configuration management server from Alibaba, originally built for the Dubbo/Spring Cloud microservices world. It handles service discovery, health checks, and dynamic config distribution — think Consul + Vault rolled into one Java monolith. Recently it has bolted on MCP server registration, A2A agent registry, and prompt/skills management, positioning itself as infrastructure for AI agent ecosystems.
Battle-tested at Alibaba scale with years of production usage across Chinese enterprises, which means the core service discovery and config propagation paths are genuinely solid. The MCP registry addition is architecturally sensible — if you're already running Nacos for microservices, reusing it as an MCP server catalog avoids running yet another registry sidecar. The Raft-based clustering for high availability is well-implemented and well-documented. First-class Kubernetes integration including DNS-based service discovery works without requiring a service mesh.
The AI features (MCP registry, A2A, prompts, skills) feel grafted on rather than designed in — the directory structure shows a hastily added `ai-registry-adaptor` module alongside the `ai` module with overlapping responsibilities, suggesting a fast-follow response to MCP hype rather than deliberate design. Documentation is split between nacos.io and inline READMEs, with a lot of it in Chinese only, which is friction for non-Chinese teams. Running it in cluster mode requires MySQL or external storage plus a Raft quorum, so 'easy to use' in the description is accurate only for standalone mode — production deployment is operationally heavy. The codebase leans hard on Spring Boot auto-wiring and convention-over-configuration, making it difficult to understand what actually happens at startup without stepping through with a debugger.