nacos-sdk-python v3

English | 简体中文

A Python implementation of Nacos OpenAPI.

see: https://nacos.io/zh-cn/docs/open-API.html

Supported Python version：

Python 3.10+

Supported Nacos version

Supported Nacos version over 3.x

Note: AI Client feature requires Nacos server version 3.1.0 or above.

Installation

 pip install nacos-sdk-python

Client Configuration

from v2.nacos import NacosNamingService, NacosConfigService, NacosAIService, ClientConfigBuilder, GRPCConfig, \
    Instance, SubscribeServiceParam, RegisterInstanceParam, DeregisterInstanceParam, \
    BatchRegisterInstanceParam, GetServiceParam, ListServiceParam, ListInstanceParam, ConfigParam
from v2.nacos.ai.model.ai_param import GetPromptParam, SubscribePromptParam, DownloadSkillParam
    
client_config = (ClientConfigBuilder()
                 .access_key(os.getenv('NACOS_ACCESS_KEY'))
                 .secret_key(os.getenv('NACOS_SECRET_KEY'))
                 .server_address(os.getenv('NACOS_SERVER_ADDR', 'localhost:8848'))
                 .log_level('INFO')
                 .grpc_config(GRPCConfig(grpc_timeout=5000))
                 .build())

• server_address - required - Nacos server address
• access_key - The aliyun accessKey to authenticate.
• secret_key - The aliyun secretKey to authenticate.
• credentials_provider - The custom access key manager.
• username - The username to authenticate.
• password - The password to authenticate.
• log_level - Log level | default: logging.INFO
• cache_dir - cache dir path. | default: ~/nacos/cache
• log_dir - log dir path. | default: ~/logs/nacos
• namespace_id - namespace id. | default: ``
• grpc_config - grpc config.
  • max_receive_message_length - max receive message length in grpc. | default: 100 * 1024 * 1024
  • max_keep_alive_ms - max keep alive ms in grpc. | default: 60 * 1000
  • initial_window_size - initial window size in grpc. | default: 10 * 1024 * 1024
  • initial_conn_window_size - initial connection window size in grpc. | default: 10 * 1024 * 1024
  • grpc_timeout - grpc timeout in milliseconds. | default: 3000
  • port_offset - gRPC port offset. The gRPC port = HTTP port + offset. | default: 1000
  • capability_negotiation_timeout - timeout for capability negotiation in milliseconds. | default: 5000
• tls_config - tls config
  • enabled - whether enable tls.
  • ca_file - ca file path.
  • cert_file - cert file path.
  • key_file - key file path.
• kms_config - aliyun kms config
  • enabled - whether enable aliyun kms.
  • endpoint - aliyun kms endpoint.
  • access_key - aliyun accessKey.
  • secret_key - aliyun secretKey.
  • password - aliyun kms password.

Config Client

config_client = await NacosConfigService.create_config_service(client_config)

config client common parameters

param: ConfigParam

• param data_id Data id.
• param group Group, use DEFAULT_GROUP if no group specified.
• param content Config content.
• param tag Config tag.
• param app_name Application name.
• param beta_ips Beta test ip address.
• param cas_md5 MD5 check code.
• param type Config type.
• param src_user Source user.
• param encrypted_data_key Encrypted data key.
• param kms_key_id Kms encrypted data key id.
• param usage_type Usage type.

Get Config

content = await config_client.get_config(ConfigParam(
            data_id=data_id,
            group=group
        ))

• param ConfigParam config client common parameters. When getting configuration, it is necessary to specify the required data_id and group in param.
• return Config content if success or an exception will be raised.

Get value of one config item following priority:

• Step 1 - Get from local failover dir.

• Step 2 - Get from one server until value is got or all servers tried.

  • Content will be saved to snapshot dir after got from server.

• Step 3 - Get from snapshot dir.

Add Listener

async def config_listener(tenant, data_id, group, content):
    print("listen, tenant:{} data_id:{} group:{} content:{}".format(tenant, data_id, group, content))

await config_client.add_listener(dataID, groupName, config_listener)

• param ConfigParam config client common parameters.
• listener listener Configure listener, defined by the namespace_id、group、data_id、content.
• return

Add Listener to a specified config item.

• Once changes or deletion of the item happened, callback functions will be invoked.
• If the item is already exists in server, callback functions will be invoked for once.
• Callback functions are invoked from current process.

Remove Listener

await client.remove_listener(dataID, groupName, config_listener)

• param ConfigParam config client common parameters.
• return True if success or an exception will be raised.

Remove watcher from specified key.

Publish Config

res = await client.publish_config(ConfigParam(
            data_id=dataID,
            group=groupName,
            content="Hello world")
        )

• param ConfigParam config client common parameters. When publishing configuration, it is necessary to specify the required data_id, group and content in param.
• return True if success or an exception will be raised.

Publish one congfig data item to Nacos.

• If the data key is not exist, create one first.
• If the data key is exist, update to the content specified.
• Content can not be set to None, if there is need to delete config item, use function remove instead.

Remove Config

res = await client.remove_config(ConfigParam(
            data_id=dataID,
            group=groupName
        ))

• param ConfigParam config client common parameters.When removing configuration, it is necessary to specify the required data_id and group in param.
• return True if success or an exception will be raised.

Remove one config data item from Nacos.

Stop Config Client

await client.shutdown()

Naming Client

naming_client = await NacosNamingService.create_naming_service(client_config)

Register Instance

response = await client.register_instance(
            request=RegisterInstanceParam(service_name='nacos.test.1', group_name='DEFAULT_GROUP', ip='1.1.1.1',
                port=7001, weight=1.0, cluster_name='c1', metadata={'a': 'b'},
                enabled=True,
                healthy=True, ephemeral=True))

Batch Register Instance

param1 = RegisterInstanceParam(service_name='nacos.test.1',
                                       group_name='DEFAULT_GROUP',
                                       ip='1.1.1.1',
                                       port=7001,
                                       weight=1.0,
                                       cluster_name='c1',
                                       metadata={'a': 'b'},
                                       enabled=True,
                                       healthy=True,
                                       ephemeral=True
                                       )
param2 = RegisterInstanceParam(service_name='nacos.test.1',
                               group_name='DEFAULT_GROUP',
                               ip='1.1.1.1',
                               port=7002,
                               weight=1.0,
                               cluster_name='c1',
                               metadata={'a': 'b'},
                               enabled=True,
                               healthy=True,
                               ephemeral=True
                               )
param3 = RegisterInstanceParam(service_name='nacos.test.1',
                               group_name='DEFAULT_GROUP',
                               ip='1.1.1.1',
                               port=7003,
                               weight=1.0,
                               cluster_name='c1',
                               metadata={'a': 'b'},
                               enabled=True,
                               healthy=False,
                               ephemeral=True
                               )
response = await client.batch_register_instances(
    request=BatchRegisterInstanceParam(service_name='nacos.test.1', group_name='DEFAULT_GROUP',
                                       instances=[param1, param2, param3]))

Deregister Instance

response = await client.deregister_instance(
          request=DeregisterInstanceParam(service_name='nacos.test.1', group_name='DEFAULT_GROUP', ip='1.1.1.1',
                                          port=7001, cluster_name='c1', ephemeral=True)
      )

Update Instance

response = await client.update_instance(
            request=RegisterInstanceParam(service_name='nacos.test.1', group_name='DEFAULT_GROUP', ip='1.1.1.1',
                                          port=7001, weight=2.0, cluster_name='c1', metadata={'a': 'b'},
                                          enabled=True,
                                          healthy=True, ephemeral=True))

Get Service

service = await client.get_service(
            GetServiceParam(service_name='nacos.test.1', group_name='DEFAULT_GROUP', cluster_name='c1'))

List Service

service_list = await client.list_services(ListServiceParam())

List Instance

instance_list = await client.list_instances(ListInstanceParam(service_name='nacos.test.1', healthy_only=True))
instance_list = await client.list_instances(ListInstanceParam(service_name='nacos.test.1', healthy_only=False))
instance_list = await client.list_instances(ListInstanceParam(service_name='nacos.test.1', healthy_only=None))

Subscribe

async def cb(instance_list: List[Instance]):
  print('received subscribe callback', str(instance_list))

await client.subscribe(
  SubscribeServiceParam(service_name='nacos.test.1', group_name='DEFAULT_GROUP', subscribe_callback=cb))

Unsubscribe

async def cb(instance_list: List[Instance]):
  print('received subscribe callback', str(instance_list))

await client.unsubscribe(
            SubscribeServiceParam(service_name='nacos.test.1', group_name='DEFAULT_GROUP', subscribe_callback=cb))

Stop Naming Client

await client.shutdown()

AI Client

Important: AI Client feature requires Nacos server version 3.1.0 or above.

from v2.nacos import NacosAIService, ClientConfigBuilder

client_config = (ClientConfigBuilder()
                 .server_address(os.getenv('NACOS_SERVER_ADDR', 'localhost:8848'))
                 .build())
                 
ai_client = await NacosAIService.create_ai_service(client_config)

Transport Modes:

• Prompt supports both gRPC and HTTP transport. By default, gRPC is used. If the gRPC port is unreachable, the AI client will still start normally (gRPC reconnects asynchronously in the background), and Prompt operations can fall back to HTTP.
• Skill download always uses HTTP, regardless of gRPC availability.
• MCP Server / Agent Card management uses gRPC.

MCP Server Management

Nacos provides management capabilities for MCP (Model Context Protocol) Server, including registration, discovery, and subscription, supporting dynamic registration and service discovery of MCP servers.

Get MCP Server

from v2.nacos.ai.model.ai_param import GetMcpServerParam

mcp_server = await ai_client.get_mcp_server(
    GetMcpServerParam(mcp_name='my-mcp-server', version='1.0.0')
)

• param GetMcpServerParam Parameter for retrieving MCP server information.
  • mcp_name - Name of the MCP server to query (required).
  • version - Version of the MCP server to query (optional).
• return McpServerDetailInfo if success or an exception will be raised.

Release MCP Server

from v2.nacos.ai.model.ai_param import ReleaseMcpServerParam
from v2.nacos.ai.model.mcp.mcp import McpServerBasicInfo, ServerVersionDetail

server_spec = McpServerBasicInfo(
    name='my-mcp-server',
    description='My MCP Server',
    protocol='http',
    versionDetail=ServerVersionDetail(version='1.0.0')
)

result = await ai_client.release_mcp_server(
    ReleaseMcpServerParam(server_spec=server_spec)
)

• param ReleaseMcpServerParam Parameter for releasing/publishing MCP server.
  • server_spec - Basic information specification for the MCP server (required).
  • tool_spec - Tool specification defining the tools provided by MCP server (optional).
  • mcp_endpoint_spec - Endpoint specification for MCP server network configuration (optional).
• return Server ID if success or an exception will be raised.

Register MCP Server Endpoint

from v2.nacos.ai.model.ai_param import RegisterMcpServerEndpointParam

await ai_client.register_mcp_server_endpoint(
    RegisterMcpServerEndpointParam(
        mcp_name='my-mcp-server',
        address='127.0.0.1',
        port=8080,
        version='1.0.0'
    )
)

• param RegisterMcpServerEndpointParam Parameter for registering MCP server endpoint.
  • mcp_name - Name of the MCP server (required).
  • address - IP address or hostname of the MCP server endpoint (required).
  • port - Port number of the MCP server endpoint (required).
  • version - Version of the MCP server (optional).

Subscribe MCP Server

from v2.nacos.ai.model.ai_param import SubscribeMcpServerParam

async def mcp_listener(mcp_id, namespace_id, mcp_name, mcp_server_detail):
    print(f"MCP Server changed: {mcp_name}, version: {mcp_server_detail.version}")

await ai_client.subscribe_mcp_server(
    SubscribeMcpServerParam(
        mcp_name='my-mcp-server',
        version='1.0.0',
        subscribe_callback=mcp_listener
    )
)

• param SubscribeMcpServerParam Parameter for subscribing to MCP server changes.
  • mcp_name - Name of the MCP server to subscribe to (required).
  • version - Version of the MCP server to subscribe to (optional).
  • subscribe_callback - Callback function to handle MCP server changes (required).

Unsubscribe MCP Server

await ai_client.unsubscribe_mcp_server(
    SubscribeMcpServerParam(
        mcp_name='my-mcp-server',
        version='1.0.0',
        subscribe_callback=mcp_listener
    )
)

Agent Card Management

Nacos provides management capabilities for AI Agent, including registration, discovery, and subscription, supporting dynamic registration and service discovery of Agent Card based on A2A protocol.

Get Agent Card

from v2.nacos.ai.model.ai_param import GetAgentCardParam

agent_card = await ai_client.get_agent_card(
    GetAgentCardParam(agent_name='my-agent', version='1.0.0')
)

• param GetAgentCardParam Parameter for retrieving agent card information.
  • agent_name - Name of the agent card (required).
  • version - Target version, if null or empty, get latest version (optional).
  • registration_type - Registration type: 'url' or 'service' (optional).
• return AgentCardDetailInfo if success or an exception will be raised.

Release Agent Card

from v2.nacos.ai.model.ai_param import ReleaseAgentCardParam
from a2a.types import AgentCard

agent_card = AgentCard(
    name='my-agent',
    version='1.0.0',
    protocol_version='1.0'
)

await ai_client.release_agent_card(
    ReleaseAgentCardParam(
        agent_card=agent_card,
        registration_type='service',
        set_as_latest=True
    )
)

• param ReleaseAgentCardParam Parameter for releasing/publishing agent card.
  • agent_card - Agent card information (required).
  • registration_type - Registration type: 'url' or 'service' (optional, default: 'service').
  • set_as_latest - Whether to set as the latest version (optional, default: False).

Register Agent Endpoint

from v2.nacos.ai.model.ai_param import RegisterAgentEndpointParam

await ai_client.register_agent_endpoint(
    RegisterAgentEndpointParam(
        agent_name='my-agent',
        address='127.0.0.1',
        port=8080,
        version='1.0.0',
        transport='JSONRPC',
        path='/agent',
        support_tls=False
    )
)

• param RegisterAgentEndpointParam Parameter for registering agent endpoint.
  • agent_name - Name of the agent (required).
  • address - IP address or hostname of the agent endpoint (required).
  • port - Port number of the agent endpoint (required).
  • version - Version of the agent (required).
  • transport - Transport protocol (optional, default: 'JSONRPC').
  • path - URL path for the endpoint (optional).
  • support_tls - Whether TLS is supported (optional, default: False).

Deregister Agent Endpoint

from v2.nacos.ai.model.ai_param import DeregisterAgentEndpointParam

await ai_client.deregister_agent_endpoint(
    DeregisterAgentEndpointParam(
        agent_name='my-agent',
        address='127.0.0.1',
        port=8080,
        version='1.0.0'
    )
)

• param DeregisterAgentEndpointParam Parameter for deregistering agent endpoint.
  • agent_name - Name of the agent (required).
  • address - IP address or hostname of the agent endpoint (required).
  • port - Port number of the agent endpoint (required).
  • version - Version of the agent (required).

Subscribe Agent Card

from v2.nacos.ai.model.ai_param import SubscribeAgentCardParam

async def agent_listener(agent_name, agent_card_detail):
    print(f"Agent card changed: {agent_name}, version: {agent_card_detail.version}")

await ai_client.subscribe_agent_card(
    SubscribeAgentCardParam(
        agent_name='my-agent',
        version='1.0.0',
        subscribe_callback=agent_listener
    )
)

• param SubscribeAgentCardParam Parameter for subscribing to agent card changes.
  • agent_name - Name of the agent (required).
  • version - Version of the agent (optional).
  • subscribe_callback - Callback function to handle agent card changes (required).

Unsubscribe Agent Card

await ai_client.unsubscribe_agent_card(
    SubscribeAgentCardParam(
        agent_name='my-agent',
        version='1.0.0',
        subscribe_callback=agent_listener
    )
)

Prompt Management

Nacos provides prompt template management capabilities, including retrieval, subscription, and rendering with variable substitution.

Get Prompt

from v2.nacos.ai.model.ai_param import GetPromptParam

prompt = await ai_client.get_prompt(
    GetPromptParam(prompt_key='my-prompt', version='1.0.0')
)
print(prompt.template)

• param GetPromptParam Parameter for retrieving prompt information.
  • prompt_key - Key of the prompt to query (required).
  • version - Version of the prompt (optional).
  • label - Label of the prompt (optional).
• return Prompt if success or an exception will be raised.

Render Prompt with Variables

The Prompt object supports template rendering with {{variableName}} placeholders. Variables defined in the prompt may include default values via PromptVariable.defaultValue. When rendering, default values are applied first, then overridden by user-provided values.

# Render the prompt template with variable substitution
result = prompt.render({"name": "Alice", "place": "Nacos"})
print(result)  # e.g. "Hello Alice, welcome to Nacos!"

# Variables with defaultValue will be used automatically if not overridden
# For example, if the prompt has a variable: PromptVariable(name="lang", defaultValue="en")
# Calling render without providing "lang" will use "en" as the value
result = prompt.render({"name": "Alice"})

• param variables - A dict of variable name to value mappings (optional). Overrides default values defined in PromptVariable.defaultValue.
• return Rendered string with all {{variableName}} placeholders replaced.

Subscribe Prompt

from v2.nacos.ai.model.ai_param import SubscribePromptParam

async def prompt_listener(prompt_key, prompt):
    print(f"Prompt changed: {prompt_key}, version: {prompt.version}")

prompt = await ai_client.subscribe_prompt(
    SubscribePromptParam(
        prompt_key='my-prompt',
        version='1.0.0',
        subscribe_callback=prompt_listener
    )
)

• param SubscribePromptParam Parameter for subscribing to prompt changes.
  • prompt_key - Key of the prompt to subscribe to (required).
  • version - Version of the prompt (optional).
  • label - Label of the prompt (optional).
  • subscribe_callback - Callback function to handle prompt changes (required).
• return Current Prompt if success or an exception will be raised.

Unsubscribe Prompt

await ai_client.unsubscribe_prompt(
    SubscribePromptParam(
        prompt_key='my-prompt',
        version='1.0.0',
        subscribe_callback=prompt_listener
    )
)

Skill Download

Nacos supports downloading skill packages as ZIP archives.

Download Skill ZIP

from v2.nacos.ai.model.ai_param import DownloadSkillParam

zip_bytes = await ai_client.download_skill_zip(
    DownloadSkillParam(skill_name='my-skill', version='1.0.0')
)

# Save to file
with open('my-skill.zip', 'wb') as f:
    f.write(zip_bytes)

• param DownloadSkillParam Parameter for downloading a skill ZIP.
  • skill_name - Name of the skill (required).
  • version - Target skill version (optional, defaults to latest).
  • label - Target skill label, e.g. "latest", "stable" (optional).
• return ZIP file content as bytes if success or an exception will be raised.

Stop AI Client

await ai_client.shutdown()