Nacos Overview

Nacos is pronounced /nɑ:kəʊs/. The name comes from Dynamic Naming and Configuration Service.

Nacos is a dynamic service discovery, configuration management, and AI Registry platform for cloud-native and AI applications. The project started from two core problems: how applications find services, and how applications safely read and update configuration. In Nacos 3.x, these capabilities are extended with AI Registry, which manages AI resources such as Skills, A2A Agents, MCP Servers, Prompts, and AgentSpecs.

The goal is simple: applications should safely and quickly find the services, configuration, and AI capabilities they need at runtime.

Core Capabilities

Capability	What it solves	Start here
Service discovery	Register, discover, subscribe to, and check services	Quick Start, Client API
Configuration management	Manage, update, roll back, and audit configuration	Quick Start, Java SDK
AI Registry	Register, govern, and discover Skills, Agents, MCP Servers, Prompts, and AgentSpecs	AI Registry Overview
Operations	Deploy, monitor, upgrade, and secure Nacos clusters	Deployment, Monitoring, Authorization
Plugins	Extend auth, data source, encryption, control, environment, and tracing behavior	Plugins

What Is New In Nacos 3.x

Nacos 3.x keeps the service discovery and configuration management capabilities, and adds stronger API, security, and AI features.

Unified v3 APIs: Client API, Admin API, and Console API serve different callers with clearer boundaries.
Stronger default security: Console and management APIs pay more attention to authentication and authorization.
AI Registry as a first-class capability: Nacos can manage MCP Servers, A2A Agents, Prompts, Skills, AgentSpecs, and related versions.
Richer plugin model: Auth, visibility, publish Pipeline, resource import, data source, and tracing can be extended when needed.
Clearer operations model: Deployment, monitoring, upgrade, system parameters, Admin API, and Maintainer SDK are easier to use in platform operations.

Product Advantages

Easy to use: Nacos provides a console, SDKs, OpenAPI, and Maintainer SDK. Developers can quickly connect applications to service discovery and configuration management. Operators can manage clusters through UI pages and APIs.
Complete capability set: Nacos covers service discovery, configuration management, health checks, configuration history, gray release, authorization, monitoring, and plugin extension. Nacos 3.x also adds AI Registry for Skill, A2A, MCP, Prompt, AgentSpec, and other AI application resources.
Production-oriented: Nacos supports cluster mode, external databases, metrics, authentication, Admin API, and upgrade workflows. It can grow from local development to production operations.
Open ecosystem: Nacos works with Spring Cloud, Dubbo, Kubernetes, Higress, Dify, Spring AI Alibaba, and other ecosystems. The plugin model also lets teams extend Nacos for their own security, storage, and governance requirements.

Design Principles

Design principles

Easy to use: Core Nacos features should be easy to adopt, understand, and operate. Users should not need to understand internal implementation details before they can register services, read configuration, or discover AI resources.
Standards-oriented: Nacos prefers clear and stable interfaces and models. Service discovery, configuration management, v3 APIs, MCP, A2A, OpenAPI, and plugin extension should avoid unnecessary private constraints.
Highly available: Nacos is runtime infrastructure. Service discovery, configuration management, and AI resource discovery can all affect running applications. Nacos therefore continues to improve clustering, storage, push, recovery, and observability.
Easy to extend: Different teams have different security, audit, data source, and release processes. Nacos uses plugins and clear API boundaries so users can extend capabilities without changing core code.

Architecture

Nacos architecture

Nacos is built on communication, consistency, storage, and runtime foundation modules. On top of these foundations, it provides service discovery, configuration management, and AI Registry.

SDKs, OpenAPI, the console, and the Maintainer SDK are the main ways users access these capabilities.

Plugins and ecosystem integrations sit around the core. Plugins extend Nacos itself. Ecosystem integrations connect Nacos with Spring Cloud, Dubbo, Kubernetes, Higress, Dify, Spring AI Alibaba, and other systems.

AI Registry

Nacos 3.x treats AI Registry as a core capability beside service discovery and configuration management. It is useful to both platform administrators and AI application developers.

AI Registry focuses on the resources that AI applications need at runtime: Skills, Agents, MCP Servers, Prompts, and AgentSpecs. These resources are not just static descriptions. They also need versions, labels, visibility, lifecycle states, and runtime discovery.

Skills package reusable AI capabilities and distribute them by version.
Agents expose AgentCards and callable endpoints for multi-agent applications.
MCP Servers expose tools, resources, and endpoints to models, Agents, and MCP clients.
Prompts provide stable templates that applications can read by version or label.
AgentSpecs distribute standardized Agent specification packages for Agent platforms and developer tools.

For the full AI Registry panorama, MCP Registry details, and resource-specific entry points, read AI Registry Overview.

Data Model

Nacos data model

Nacos resources are usually isolated by namespace.

In service discovery, the resource name is usually the service name.
In configuration management, the resource name is usually the Data ID.
In AI Registry, the resource name can be an MCP Server name, Agent name, Prompt key, Skill name, or AgentSpec name.

Use namespaces to separate environments, tenants, or business domains. In production, avoid mixing test and production resources in the same namespace.

Deployment Modes

Nacos deployment modes

Nacos supports standalone mode and cluster mode.

Standalone mode is suitable for local development, feature validation, and test environments. It is easy to start and can use the embedded Derby database. It is not recommended for highly available production traffic.

Cluster mode is suitable for production. Multiple nodes provide service together. With an external database and consistency protocols, the cluster has better availability. For production, also configure monitoring, alerting, authentication, and backup.

Ecosystem

Nacos ecosystem

Nacos ecosystem integrations cover microservice, cloud-native, and AI application scenarios. Common integrations include Spring Cloud, Dubbo, Kubernetes, CoreDNS, Higress, Dify, Spring AI Alibaba, and Nacos MCP Router.

Ecosystem components usually serve two purposes. Some connect applications to Nacos, such as Spring Cloud, Dubbo, and SDKs. Others bring Nacos registration, configuration, or AI Registry capabilities into larger platforms, such as Kubernetes, Higress, Dify, and MCP Router.

Roadmap

Nacos roadmap

Nacos continues to evolve in three directions: stable service discovery and configuration management, production-grade security and operations, and AI Registry governance for AI applications. The roadmap may change with community feedback and release planning. Use release notes and milestones as the source of truth for delivered features.

Next Actions

Try Nacos for the first time: read Quick Start.
Prepare for production: read Deployment.
Use AI Registry: read AI Registry Overview.
Understand API boundaries: read API Overview.
Contribute to the community: read Contributing.