How to Seamlessly Integrate Dify with Nacos MCP for Dynamic AI Service Discovery and Prompt Management

Problem Statement

Dify is an open‑source LLM application platform that lacks native Model Context Protocol (MCP) support, requires workflow redeployment for prompt changes, and has complex deployment configuration across many components.

Nacos 3.0 Capabilities

MCP dynamic management : Nacos 3.0 can automatically register various MCP server types, provides a standard MCP Registry API, and supports private deployment of the registry.

Runtime prompt updates : Prompts stored in Nacos can be fetched by Dify at runtime, eliminating the need to rebuild workflows.

Centralized environment variables : Dify can read .env parameters from Nacos, enabling white‑screen configuration changes and version history.

Step‑by‑Step Integration

1. Register MCP Services in Nacos

Manually add a service via the Nacos console.

Use the MCP auto‑register guide (https://nacos.io/docs/latest/manual/user/ai/mcp-auto-register/).

Convert existing HTTP APIs to MCP services with zero code changes (https://nacos.io/docs/latest/manual/user/ai/api-to-mcp/).

After registration the MCP server appears in the Nacos UI.

2. Install and Configure the Nacos MCP Plugin in Dify

Search for "Nacos MCP" in the Dify plugin marketplace, install it, and configure the following fields:

SERVER_ADDR=<nacos_address>
USERNAME=<username>
PASSWORD=<password>
NAMESPACE=<namespace>
DATA_ID=<data_id>
GROUP=<group>

3. Use the Plugin in a Dify Workflow

Add a tool node that calls the Nacos MCP plugin functions: list_mcp_servers – returns all registered MCP servers. list_mcp_server_tools – returns the tool list of a selected server. list_mcp_server_tools_by_user – returns tools configured for a specific user.

The model can dynamically select the appropriate MCP service during execution.

4. Dynamic Prompt Management

Create a configuration file in Nacos (e.g., Prompt.template) containing the prompt text. In the Dify workflow, add a "Read Nacos" node to fetch the prompt at runtime and feed it to the Agent node. Updating the prompt in Nacos instantly affects the running application without redeployment.

5. Centralized Environment Variable Hosting

Set the following environment variables when launching Dify to enable Nacos as the remote settings source:

REMOTE_SETTINGS_SOURCE_NAME=nacos
DIFY_ENV_NACOS_SERVER_ADDR=<nacos_addr>
DIFY_ENV_NACOS_DATA_ID=<data_id>
DIFY_ENV_NACOS_GROUP=<group>
DIFY_ENV_NACOS_NAMESPACE=<namespace>
DIFY_ENV_NACOS_USERNAME=<username>
DIFY_ENV_NACOS_PASSWORD=<password>
DIFY_ENV_NACOS_ACCESS_KEY=<access_key>
DIFY_ENV_NACOS_SECRET_KEY=<secret_key>

Dify reads its startup parameters from Nacos, allowing white‑screen changes and version rollback.

Benefits

{Rapid integration of new MCP services without code changes.}

{Reduced token consumption by letting the model query only necessary tools.}

{Instant prompt iteration without redeploying workflows.}

{Centralized configuration and version history simplify operations.}

This integration demonstrates a cloud‑native approach to building, deploying, and maintaining generative AI applications with high agility and low operational overhead.