Supported Modules

After the agent is loaded, it automatically instruments the packages below — no code changes needed. This page reproduces the support table from the repository's compatibility.md for version 14.13.0 of the agent, plus the Node.js core modules and the AI platform matrix.

Note: "Latest published version" is the newest version of the package at the time the table was generated; it may not be the newest version the agent supports.

Requirements

  • Node.js 22.16.0 or later (engines.node is >=22.16.0). On an older version the agent logs New Relic for Node.js requires a version of Node >=22.16.0 ... Not starting! and stays disabled; on Node.js 27 or later it starts but logs a warning that the version is not yet supported.
  • Packages must be loaded after the agent (node -r newrelic or require('newrelic') first). ES module apps also need the ESM loader — see Loading the Agent.

Node.js Core Modules

The agent instruments these built-in modules (subscribers in lib/subscribers/core/):

Module What is recorded
http, https Inbound requests (web transactions) and outbound requests (external segments, distributed tracing headers).
http2 Outbound HTTP/2 client sessions (connect, request).
child_process Segments for exec and execFile.
crypto Segments for pbkdf2, randomBytes, pseudoRandomBytes, randomFill, and scrypt.
dns Segments for DNS lookups.
fs Segments for file system operations.
net Outbound sockets (connect, createConnection).
zlib Segments for asynchronous compression.
process (globals) uncaughtException and unhandledRejection are reported as errors.

Third-Party Packages

Versions are the minimum supported, the latest published at the time of writing, and the agent version that introduced support.

Package name Minimum supported version Latest published version Introduced in*
@anthropic-ai/sdk 0.33.0 1.8.0 13.19.0
@apollo/gateway 2.3.0 2.14.4 14.0.0
@apollo/server 4.0.0 5.5.2 14.0.0
@aws-sdk/client-bedrock-runtime 3.377.0 3.1169.0 11.13.0
@aws-sdk/client-dynamodb 3.377.0 3.1169.0 8.7.1
@aws-sdk/client-sns 3.377.0 3.1169.0 8.7.1
@aws-sdk/client-sqs 3.377.0 3.1169.0 8.7.1
@aws-sdk/lib-dynamodb 3.377.0 3.1169.0 8.7.1
@aws-sdk/smithy-client 3.47.0 3.374.0 8.7.1
@azure/functions 4.7.0 4.17.1 12.18.0
@elastic/elasticsearch 7.16.0 9.5.1 11.9.0
@google/adk 1.1.0 2.5.0 13.20.0
@google/genai 1.1.0 3.3.0 12.21.0
@grpc/grpc-js 1.4.0 1.14.7 8.17.0
@hapi/hapi 20.1.2 21.4.13 9.0.0
@hapi/vision 5.0.0 7.0.3 9.0.0
@koa/router 12.0.1 15.7.1 3.2.0
@langchain/aws 0.1.3 1.5.2 13.8.0
@langchain/core 0.2.0 1.4.1 11.13.0
@langchain/langgraph 1.0.0 1.7.4 13.12.0
@langchain/openai 0.2.0 1.8.4 11.13.0
@modelcontextprotocol/sdk 1.13.0 1.32.2 13.2.0
@nestjs/core 10.0.0 12.5.0 10.1.0
@opensearch-project/opensearch 2.1.0 3.10.0 12.10.0
@prisma/client 5.0.0 7.10.1 11.0.0
@smithy/smithy-client 2.0.0 4.15.2 11.0.0
amqplib 0.5.0 2.4.0 2.0.0
aws-sdk 2.2.48 2.1693.0 6.2.0
bluebird 3.0.0 3.7.2 1.27.0
bunyan 1.8.12 1.8.15 9.3.0
cassandra-driver 4.0.0 4.10.0 1.7.1
connect 3.0.0 3.7.0 2.6.0
express 4.15.0 5.3.0 2.6.0
fastify 4.0.0 5.12.8 8.5.0
generic-pool 3.0.0 3.9.0 0.9.0
ioredis 4.0.0 6.0.0 1.26.2
iovalkey 0.1.0 0.4.0 13.9.0
kafkajs 2.0.0 2.2.4 11.19.0
koa 2.0.0 3.3.0 3.2.0
memcached 2.2.0 2.2.2 1.26.2
mongodb 4.1.4 7.9.0 1.32.0
mysql 2.16.0 2.18.1 1.32.0
mysql2 3.0.0 3.24.6 1.32.0
next 14.0.0 16.4.6 12.0.0
openai 4.0.0 7.57.0 11.13.0
pg 8.2.0 8.23.1 9.0.0
pg-native 3.0.0 3.9.1 9.0.0
pino 8.0.0 10.4.0 8.11.0
q 1.3.0 1.5.1 1.26.2
redis 3.1.0 6.3.1 1.31.0
restify 11.0.0 12.1.0 2.6.0
superagent 3.0.0 10.4.2 4.9.0
undici 5.0.0 8.12.4 11.1.0
when 3.7.0 3.7.8 1.26.2
winston 3.0.0 3.19.0 8.11.0

* When a package is not specified, support is built into the newrelic package.

Packages behind switches

  • kafkajs is only instrumented when the kafkajs_instrumentation feature flag is true (default false): feature_flag: { kafkajs_instrumentation: true } or NEW_RELIC_FEATURE_FLAG_KAFKAJS_INSTRUMENTATION=true.
  • AI packages (openai, @anthropic-ai/sdk, @aws-sdk/client-bedrock-runtime, @google/genai, @google/adk, LangChain, LangGraph, MCP SDK) record AI monitoring data only when AI monitoring is enabled — see AI Monitoring.
  • @prisma/client uses the optional dependency @prisma/prisma-fmt-wasm to parse the schema.
  • Each package's instrumentation can be turned off individually with its instrumentation.<package>.enabled setting — see instrumentation settings.

AI Platforms

The agent supports the following AI platforms and integrations.

Amazon Bedrock

Through the @aws-sdk/client-bedrock-runtime module, we support all models (text-only) through the Converse API; for InvokeModel API, we support the following models:

Model Image Text Vision
Amazon Titan ❌ ✅ -
Anthropic Claude ❌ ✅ ❌
Cohere ❌ ✅ -
Meta Llama3 ❌ ✅ -

Note: If a model supports streaming, we also instrument the streaming variant.

Google ADK

Through the @google/adk module, we support the following features:

Agents Tools
✅ ✅

Models/providers are generally supported transitively by our instrumentation of the provider's module.

Provider Supported Transitively
Google Gemini ✅ ✅

LangChain

The following general features of LangChain are supported:

Agents via LangGraph Chains Tools Vectorstores
✅ ✅ ✅ ✅

Models/providers are generally supported transitively by our instrumentation of the provider's module.

Provider Supported Transitively
Azure OpenAI ❌ ❌
Amazon Bedrock ✅ ✅
OpenAI ✅ ✅

Model Context Protocol SDK

Through the @modelcontextprotocol/sdk module, we support:

Prompt Retrieval Resource Reading Tool Calls
✅ ✅ ✅

MCP is provider-agnostic. Any LLM provider that uses MCP tools is supported transitively.

Anthropic

Through the @anthropic-ai/sdk module, we support:

Chat Completions Files
✅ ✅ ❌

Google Gen AI

Through the @google/genai module, we support:

Audio Cache Chat Embeddings Image PDF Text Video
❌ ❌ ✅ ✅ ❌ ❌ ✅ ❌

OpenAI

Through the openai module, we support:

Audio Chat Completions Embeddings Files Images
❌ ✅ ✅ ✅ ❌ ❌

Optional Dependencies

These are installed when available and enable extra features:

Package Used for
@datadog/pprof Profiling (profiling.* settings).
@newrelic/fn-inspect Code-level metrics (function name, file path, and line on segments).
@prisma/prisma-fmt-wasm Prisma instrumentation.

Libraries Without Instrumentation

For a library that is not in the table, either use the transaction and segment API or write custom instrumentation.