🔭 autotel
Write once, observe everywhere.
Instrument your Node.js code once and stream traces, metrics, logs, and product events to OTLP-compatible backend. No vendor lock-in.
에이전트 강화
게시자의 자체 명령입니다. PlugMCP 엔드포인트가 아닙니다.
Host a copy with our keys and logs.
npx -y [email protected]Claude Code에는 로컬 명령용 설치 링크가 없습니다. 아래 명령을 복사해 터미널에서 실행합니다.
claude mcp add autotel-mcp-instrumentation -- npx -y autotel-mcp-instrumentation@60.1.0터미널에서 실행합니다.
{ "mcpServers": { "autotel-mcp-instrumentation": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}claude_desktop_config.json
원격 URL인가요, 내 컴퓨터에서 실행하는 명령인가요? 두 설치 방식의 차이
PlugMCP에서 실행
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
실행One init(), wrap functions with trace(), and get automatic traces, metrics, and events:
import { init, trace, track } from 'autotel';
import { SlackSubscriber } from 'autotel-subscribers';
import { PostHogSubscriber } from 'autotel-posthog/subscriber';
// Initialize once at startup
init({
service: 'checkout-api',
devtools: true, // Local traces + metrics + logs in autotel-devtools
subscribers: [
new PostHogSubscriber({ apiKey: process.env.POSTHOG_KEY! }),
new SlackSubscriber({ webhookUrl: process.env.SLACK_WEBHOOK! }),
],
});
// Wrap any function - automatic spans, error tracking, and context
export const processOrder = trace(async function processOrder(
orderId: string,
amount: number,
) {
const user = await db.users.findById(orderId);
const payment = await chargeCard(user.cardId, amount);
// Product events automatically enriched with trace context
// Sent to: OTLP + PostHog + Slack (all in one call!)
track('order.completed', { orderId, amount, userId: user.id });
return payment;
});
That's it. Every call to processOrder() now:
traceId and spanId to all adaptersWrap a handler and autotel emits one canonical wide event per request (full context in a single log line) on top of real distributed traces and metrics. You get the debuggability of one event per operation and the call graph of OpenTelemetry from the same instrumentation.
autotel ships 45 Agent Skills for AI assistants (Claude Code, Cursor, Windsurf, Continue, …), organized under skills/ by core/, frameworks/, integrations/, and extending/ (build your own subscriber, exporter, or framework adapter). Add one, or the whole catalog, with the skills CLI:
npx skills add jagreehal/autotel # browse and pick
npx skills add jagreehal/autotel --skill autotel-core
npx skills add jagreehal/autotel --skill autotel-tanstack
Highlights:
review-otel-patterns: audit a codebase for OTel anti-patterns. Covers Next.js, Nuxt, Nitro, TanStack Start, Hono, Express, Fastify, Elysia, NestJS, Cloudflare Workers, AWS Lambda, edge runtimes, and standalone Node.analyze-traces: read OTLP traces from any backend, local dump, or in-memory exporter to debug failures, latency, and cardinality issues.migrate-to-autotel: cut over from raw OTel SDK, Sentry, Datadog APM, New Relic, Honeycomb Beelines, or OpenTracing.tune-sampling: head and tail sampling strategies, AI-aware and Cloudflare-aware.build-audit-trails: tamper-aware audit logs on top of OTel spans.debug-missing-spans: top-to-bottom troubleshooting walkthrough.See skills/README.md for the full index.
This monorepo contains the following packages:
Core library providing ergonomic OpenTelemetry instrumentation with:
trace(), span(), and decoratorsdefineErrorCatalog, defineAuditCatalog)GenAI/LLM observability lives in the separate autotel-genai package: canonical gen_ai.* tracing (traceGenAI), per-model cost estimation (recordLLMCost from autotel-genai/cost), metrics, events, an AI-SDK bridge, and agent governance (autotel-genai/agent).
Product events subscribers for:
Edge runtime support for:
Composable framework DX adapters on top of autotel core:
useLogger(...) style ergonomics for framework handlerswithAutotel(...) wrappers for Next, Nitro, Cloudflare, Express, and FastifyautoEmit, default on; opt out per handler)parseError() and drain pipeline compositionAudit-focused helpers for compliance logging with automatic tail-sampling bypass:
withAudit(...): structured audit metadata with automatic outcome taggingforceKeepAuditEvent(...): keep critical audit trails past tail-drop samplingsetAuditAttributes(...): normalized audit.* span attributesMigration Guide - Migrate from vanilla OpenTelemetry to autotel:
Typical migration: Replace NODE_OPTIONS and 30+ lines of SDK boilerplate with init(), wrap functions with trace() instead of manual span.start()/span.end().
npm install autotel
# Optional: Add event subscribers (PostHog, Slack, Mixpanel, etc.)
npm install autotel-subscribers
# Optional but recommended for local DX
npm install -D autotel-devtools
# or
pnpm add autotel
pnpm add autotel-subscribers # Optional
pnpm add -D autotel-devtools # Optional but recommended
For the fastest feedback loop, run local devtools and point autotel at it:
import { init, trace } from 'autotel';
init({
service: 'my-app',
devtools: true,
});
const result = await trace(async () => 'success')();
That gives you:
init() surface you can later point at Grafana, Datadog, or any OTLP backendIf you want autotel to boot the local devtools server for you:
init({
service: 'my-app',
devtools: { embedded: true },
});
This requires autotel-devtools to be installed. If it is not installed, autotel falls back to http://127.0.0.1:4318.
See traces during development without configuring a backend:
import { init, trace } from 'autotel';
// Start with console-only (no backend needed)
init({
service: 'my-app',
debug: true, // Outputs spans to console
});
// Your traced functions work as normal
const result = await trace(async () => {
// Your code here
return 'success';
})();
// Span printed to console automatically!
How it works:
debug: true - Print spans to console AND send to backend (if endpoint configured)
Or use environment variable:
AUTOTEL_DEBUG=true node server.js
Configure autotel using standard OpenTelemetry environment variables:
export OTEL_SERVICE_NAME=my-app
export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318
export OTEL_EXPORTER_OTLP_HEADERS=x-honeycomb-team=YOUR_API_KEY
export OTEL_RESOURCE_ATTRIBUTES=deployment.environment=production
For local development, devtools: true is usually a better default than setting OTEL_EXPORTER_OTLP_ENDPOINT manually.
Then call init() without any config - it picks up env vars automatically:
init({ service: 'my-app' }); // Minimal config, env vars fill the rest
# Clone and install dependencies
git clone https://github.com/jagreehal/autotel.git
cd autotel
pnpm install
# Build all packages
pnpm build
# Run tests
pnpm test
# Run example apps
pnpm --filter @jagreehal/example-basic start
pnpm --filter @jagreehal/example-http start
autotel/
├── packages/
│ ├── autotel/ # Core library
│ ├── autotel-subscribers/ # Event subscribers
│ └── autotel-edge/ # Edge runtime support
├── apps/
│ ├── example-basic/ # Basic usage example
│ ├── example-http/ # Express server example
│ └── cloudflare-example/ # Cloudflare Workers example
└── turbo.json # Turborepo configuration
# Development
pnpm dev # Watch mode for all packages
pnpm build # Build all packages
pnpm test # Run all tests
pnpm test:integration # Run integration tests
# Code quality
pnpm lint # Lint all packages
pnpm format # Format code with Prettier
pnpm type-check # TypeScript type checking
# Releases
pnpm changeset # Create a changeset
pnpm version-packages # Version packages
pnpm release # Publish to npm
pnpm --filter @jagreehal/example-basic start
pnpm --filter @jagreehal/example-http start
pnpm --filter cloudflare-example dev
We welcome contributions! Please see our contributing guidelines for details.
git checkout -b feature/my-featurepnpm testpnpm changesetgit commit -am "Add new feature"git push origin feature/my-featureWe use changesets for version management:
pnpm changeset
Follow the prompts to:
Autotel is built on top of OpenTelemetry and provides:
Having issues seeing your traces? Use ConsoleSpanExporter for visual debugging or InMemorySpanExporter for testing. See the full troubleshooting guide in the detailed docs.
"autotel" is a trademark of Jag Reehal and is not granted for use by the software licence. See TRADEMARKS.md for what you can and cannot do with the name.
OpenTelemetry instrumentation for Model Context Protocol (MCP) with distributed tracing support
{ "mcpServers": { "autotel-mcp-instrumentation": { "command": "npx", "args": [ "-y", "[email protected]" ] } }}도구
아직 도구 목록이 없습니다. 원격 서버는 첫 상태 확인 후에 도구 목록이 생깁니다.
FAQ
이 페이지에서 Cursor, Claude Desktop, Claude Code 또는 VS Code용 stdio 설정을 복사합니다. PlugMCP 프록시가 아니라 게시자의 설치 방식을 사용합니다.
이 레코드에는 필수 환경 변수가 등록되어 있지 않습니다.
Autotel MCP Instrumentation은(는) stdio를 통해 로컬에서 실행됩니다.
마지막 확인: 2026-11-15. OpenTelemetry instrumentation for Model Context Protocol (MCP) with distributed tracing support