Node.js components and frameworks supported by ARMS Application Monitoring - Application Real-Time Monitoring Service

Prerequisites

(Recommended) Node.js 14.x or later. For AI-related components and ESM Loader scenarios, use Node.js 16 or later. For production environments, use Node.js 18 or Node.js 20 LTS.
The ARMS Node.js agent is installed.
```
npm install @loongsuite/cms_node_sdk
```
The agent must load before your application frameworks and components. If modules such as express, mysql2, or openai load before the agent starts, instrumentation may be incomplete. For CommonJS applications, use the -r flag to preload the agent. For ESM applications, use the Loader approach.
```
node -r @loongsuite/cms_node_sdk/register app.js
node --experimental-loader=@loongsuite/cms_node_sdk/import-hooks app.mjs
```

Version information

The supported version ranges for Application Monitoring in the following tables are based on the automatic instrumentation component definitions in @loongsuite/cms_node_sdk@1.0.10 and its dependency packages.

For Node.js built-in modules, runtime metrics, and diagnostics channel-based instrumentation, the agent does not filter by npm package version. Instead, instrumentation depends on the Node.js runtime version and whether the module exposes a collectible entry point.

If a framework is not listed below but uses Node.js built-in modules such as http, https, undici, or net, the agent can still collect basic entry or exit traces. Framework-specific capabilities such as routing, middleware, Agent, and Tool require dedicated instrumentation.

Web frameworks

Component or framework	npm package	Supported versions	Collected data
Express	express	>=4.0.0 <6.0.0	HTTP entry traces, routing information, middleware and handler execution, and error messages
Koa	koa	>=2.0.0 <4.0.0	HTTP entry traces, middleware execution, routing information, and error messages
Next.js	next	>=12.0.0 <17.0.0	Server-side requests, SSR, RSC, API routes, error pages, static pages, routing information, status codes, latency, and error messages
Nuxt 2	@nuxt/server, @nuxt/vue-renderer, @nuxt/core	>=2.3.0 <3.0.0	SSR, SPA, API, server middleware, `asyncData`, `fetch`, `validate`, and Vue renderer stages
Nuxt 3/Nitro/h3	h3, hookable, Node.js `http`/`https`	h3 >=1.0.0 <2.0.0; hookable >=5.0.0 <6.0.0	Server-side requests, h3 handlers, Nitro hooks, and Node.js request listeners
Socket.IO	socket.io	2.x, >=3.0.0 <5.0.0	Event sending and receiving, namespaces, room broadcasting, message latency, and error messages

HTTP, RPC, and network communication

Component or module	npm package	Supported versions	Collected data
Node.js HTTP	http, https	Node.js built-in module	HTTP server entries, HTTP client exits, status codes, request latency, request and response sizes, and error messages
Undici	undici	No fixed semver range	HTTP client requests, status codes, latency, and error messages collected through the diagnostics channel
gRPC	@grpc/grpc-js	>=1.0.0 <2.0.0	gRPC client and server calls, service and method, status codes, latency, and error messages
DNS	dns, dns/promises	Node.js built-in module	DNS lookup calls, resolution latency, and error messages. DNS instrumentation is disabled by default starting with version 1.0.9. Explicitly enable it with `dns: { enabled: true }` when needed.
TCP/Unix Socket	net	Node.js built-in module	TCP and Unix Socket connections, destination addresses, ports, and connection errors

Databases and ORMs

Component or framework	npm package	Supported versions	Collected data
MySQL	mysql	>=2.0.0 <3.0.0	SQL call traces, SQL statements, connection information, slow queries, and error messages
MySQL2	mysql2	>=1.4.2 <4.0.0	SQL call traces, query and execute operations, SQL statements, connection information, slow queries, and error messages
PostgreSQL	pg	>=8.0.3 <9.0.0	SQL call traces, SQL statements, connection information, slow queries, and error messages
PostgreSQL Pool	pg-pool	>=2.0.0 <4.0.0	Connection pool connect calls and supplementary database access trace information
MongoDB	mongodb	>=3.3.0 <4.0.0, >=4.0.0 <7.0.0	MongoDB commands, collections, database names, connection information, latency, and error messages
Mongoose	mongoose	>=5.9.7 <9.0.0	Model, Query, and Aggregate calls, collection information, database operations, latency, and error messages

Caching and message queues

Component or framework	npm package	Supported versions	Collected data
Redis	redis	>=2.6.0 <4.0.0	Redis commands, connection information, call latency, and error messages
Node Redis Client	@redis/client, @node-redis/client	>=1.0.0 <2.0.0, >=5.0.0 <6.0.0	Redis v4 and v5 client commands, multi commands, connection information, call latency, and error messages
ioredis	ioredis	>=2.0.0 <6.0.0	Redis commands, connection information, call latency, and error messages
KafkaJS	kafkajs	>=0.3.0 <3.0.0	Kafka producer and consumer calls, topics, message send and receive latency, and error messages

Logging frameworks

Component or framework	npm package	Supported versions	Collected data
Pino	pino	>=7.0.0 <11.0.0	Trace correlation field injection into logs
Winston	winston	>=3.0.0 <4.0.0	Trace correlation field injection into logs
Bunyan	bunyan	>=1.8.0 <3.0.0	Trace correlation field injection into logs
Console	console	Node.js global object	Trace correlation field injection into console.log, console.info, console.warn, and console.error

AI application components

Component or framework	npm package	Supported versions	Collected data
OpenAI Node SDK	openai	>=4.0.0 <6.0.0	Chat, Completion, and Image calls, model information, token usage, streaming calls, and error messages
LangChain JS	@langchain/core, langchain	No fixed semver range. Subject to the availability of module entry points such as @langchain/core/callbacks/manager.	Chain, LLM, Tool, Retriever, and Embedding run traces, token usage, and error messages
LangGraph	@langchain/langgraph	No fixed semver range. Subject to the availability of module entry points such as @langchain/langgraph/prebuilt and @langchain/langgraph/pregel.	ReAct Agent, Pregel stream, Agent and Step metadata, forming complete Agent traces in collaboration with LangChain instrumentation
Vercel AI SDK	ai	>=4.0.0 <7.0.0	generateText, streamText, generateObject, streamObject, Tool, Embedding, Rerank, and Agent call traces
Anthropic Claude Agent SDK	@anthropic-ai/claude-agent-sdk	>=0.3.0 <1.0.0	query and startup calls, Agent runs, model requests, tool calls, session information, token usage, and error messages
Mastra	@mastra/core, @mastra/observability	@mastra/core >=1.50.0 && <2.0.0	Agent, Step, LLM, Tool, and Workflow spans, plus input and output, tokens, errors, and RED metrics. Install `@mastra/observability`.
Genkit	@genkit-ai/core	>=1.0.4 && <2.0.0	Flow, Agent, Model, Tool, Retriever, and Embedding call traces, token usage, and error messages
MCP SDK v1	@modelcontextprotocol/sdk	>=1.0.0 && <2.0.0	MCP Server request, notification, and tool handler spans, including parameters, results, resource URIs, JSON-RPC errors, and latency
MCP Server v2	@modelcontextprotocol/server	>=2.0.0 && <3.0.0	Automatic collection of MCP v2 Server requests, notifications, and tool handler spans
Chrome DevTools MCP	chrome-devtools-mcp	>=1.8.0 && <1.9.0	Chrome DevTools tool calling traces. `tools/call` is reported as a TOOL span in GenAI traces and metrics.
Pi Agent SDK	@earendil-works/pi-agent-core, @earendil-works/pi-ai, @earendil-works/pi-coding-agent, @earendil-works/pi-durable	>=0.74.0 <2; pi-durable >=1.2.0 <1.3.0	Agent, Step, LLM, Tool, compaction, tree navigation, streaming responses, retries, and queue events. Collects model, provider, token, cached token, cost, time to first token, status, and duration.
OpenClaw	openclaw, @loongsuite/cms_instrumentations/openclaw	Added in version 1.0.11. Validated with OpenClaw 2026.10.5; no complete supported-version range is defined.	Collects model, tool, sub-agent, error, and cancellation traces. In RUM scenarios, content is not collected by default; this default does not apply to a standard NodeSDK. The validated environment includes standalone Node agents and gateways, plus main, worker, utility, spawn, and fork processes in Electron 44.9.1. ASAR validation covers runtime extraction and injection only; it does not constitute complete OpenClaw ASAR business validation.

Node.js runtime metrics

Metric category	Collection target	Collected data
Event Loop	Node.js event loop	Event loop delay, event loop utilization, and other metrics
V8	V8 engine	Heap memory, GC count, GC latency, heap space usage, and other metrics
Process	Node.js process	CPU, memory, thread-related metrics, open file descriptors, context switches, virtual memory, and other process metrics

Additional integration capabilities

Integration capability	Description
RUM Electron	Added in version 1.0.11. Provides a Node.js Trace runtime that can accept callbacks or handoff through an authenticated local HTTP channel for use with RUM Electron.
Open-source OpenTelemetry instrumentation	Uses the OpenTelemetry API compatibility bridge to bring traces, metrics, and context from open-source instrumentation into NodeSDK. It automatically deduplicates overlapping NodeSDK native collection. Install and enable third-party instrumentation separately; actual support depends on validation.
`genai_utils` compatibility integration	Uses the OpenTelemetry API compatibility bridge to bring GenAI spans created by `genai_utils` into NodeSDK trace collection and reporting. The spans share context with NodeSDK automatic collection to associate Agent, model, and tool calls. Install and initialize `genai_utils` in the application.

Configuration

Most built-in instrumentations are enabled by default. DNS instrumentation is disabled by default starting with version 1.0.9. Explicitly enable it with dns: { enabled: true } when needed. You can adjust instrumentation per application in the ARMS console or through the instrumentations field when initializing the SDK.

The following example configures the instrumentations field:

const { NodeSDK } = require('@loongsuite/cms_node_sdk');

const sdk = new NodeSDK({
  serviceName: 'your-app-name',
  licenseKey: 'your-license-key',
  regionId: 'cn-hangzhou',
  instrumentations: {
    next: {
      enabled: true
    },
    nuxt: {
      enabled: true
    },
    mysql2: {
      enabled: true
    },
    openai: {
      enabled: true
    },
    vercel_ai: {
      enabled: true
    },
    mastra: { enabled: true },
    genkit: { enabled: true },
    mcp: { enabled: true },
  },
});

sdk.start();

Common configuration keys:

Category	Configuration keys
HTTP and network	`http`, `undici`, `grpc`, `dns`, `net`
Web frameworks	`express`, `koa`, `next`, `nuxt`, `socketio`
Databases	`mysql`, `mysql2`, `pg`, `mongodb`, `moogoose`*
Caching and messaging	`redis`, `ioredis`, `kafka`
Logging	`pino`, `winston`, `bunyan`, `console`
AI	`openai`, `langchain`, `langgraph`, `vercel_ai`, `claude_agent`, `mastra`, `genkit`, `mcp`, `openclaw`, `pi_agent`
Runtime	`runtime`
Top-level SDK	`processIsolation` (optional identity isolation, disabled by default). When enabled, initialize the SDK separately in each business process or worker so Trace, Metric, and Log identities align. Explicit `service.instance.id` values are retained. `host.ip` and `net.host.ip`/`net.host.ipv4` are fixed to the actual network interface IP address and override custom logical IP addresses.

Note

The configuration key for Mongoose is moogoose, not mongoose. This is a known code-level key name. The Mongoose component is displayed as mongoose in user-facing contexts. If you explicitly configure the Mongoose instrumentation switch in code, use moogoose.

Usage notes

Automatic instrumentation requires the agent to load before target modules. If your application loads modules such as express, next, nuxt, mysql2, openai, @mastra/core, @genkit-ai/core, or MCP Server modules before starting the agent, instrumentation for those components may be incomplete.
Next.js automatic instrumentation is supported starting with @loongsuite/cms_node_sdk@1.0.5. Nuxt 2 automatic instrumentation is supported starting with @loongsuite/cms_node_sdk@1.0.7. Mastra, Genkit, and MCP Server automatic instrumentation are supported starting with version 1.0.9.
Nuxt 2 instrumentation uses the @nuxt/server, @nuxt/vue-renderer, and @nuxt/core modules. Nuxt versions earlier than 2.3.0 are not supported.
AI components can involve content such as prompts, responses, and tool inputs and outputs. MCP collects tool parameters, parameter summaries, and results, using a fixed-length safe JSON projection to limit attribute size. Screenshots, audio, blobs, and other binary content are omitted. MCP does not currently provide separate redaction settings. Before enabling it in a production environment, confirm that this content is permitted in traces.
MCP instrumentation is enabled by default in version 1.0.9. The initial release provides only the mcp: { enabled: boolean } switch and supports the server side only. MCP SDK v1 requires Node.js 18 or later, and MCP Server v2 requires Node.js 20 or later. chrome-devtools-mcp 1.8.x requires Node.js 20.19+, 22.12+, or 23+. ESM applications continue to use the existing import-hooks Loader and do not require an additional registration entry point.
MCP tools/call operations are marked as gen_ai.span.kind=TOOL. Protocol operations such as resource and prompt remain ordinary MCP Server spans and are not inferred as Retriever or Reranker spans by name. Actual retrieval or reranking calls inside a tool must produce nested spans through the relevant framework or model SDK instrumentation.
Starting with version 1.0.9, the agent sends its INFO, DEBUG, WARN, and ERROR diagnostic output to stderr according to the configured log level. Application console.log output remains on stdout. Applications using a stdio MCP Server must ensure that business logs are not written to stdout.
For unlisted components, the agent may still collect basic traces through underlying HTTP, RPC, or database client instrumentation, but framework-specific semantic fields and view capabilities are not guaranteed.
Component version ranges are updated with each agent release. After upgrading, check the documentation and release notes for updated version support.