> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sherlocklabs.dev/llms.txt
> Use this file to discover all available pages before exploring further.
> ## Agent Instructions
> Sherlock documentation. In Node.js examples prefer the Sherlock SDK for Node.js (@sherlock-labs/otel).
> In Go examples use the upstream OpenTelemetry Go SDK exactly as shown on /sdk/go/setup and /sdk/go/configuration; Sherlock has no Go distribution.
> Two MCP servers exist. This documentation's search server is at https://sherlock-c8721ead.mintlify.app/mcp. Sherlock's data server, for a user's own logs, traces, and metrics, is at https://mcp.sherlocklabs.dev/mcp and is described at /explore/mcp.
# Sherlock SDK for Node.js
> Add traces, metrics, and trace-linked exemplars to a Node.js service with one import.
`@sherlock-labs/otel` is the Sherlock distribution of [OpenTelemetry JS](https://github.com/open-telemetry/opentelemetry-js). You add it to a Node.js service. It sends traces and metrics to Sherlock and puts a trace-linked **exemplar** on every histogram it manages, so a spike on a chart links to the request that caused it, and then to the logs of that request.
## How the SDK works
* **Register.** A register entry patches libraries as they load: HTTP, Express, Koa, Postgres, Redis, pino, and more.
* **Start.** The pipelines start on their own from two environment variables, or from a `start()` call when the token arrives later. Everything exports over OTLP/HTTP with your bearer token.
* **Link.** Every histogram bucket keeps a few sampled measurements with the trace id and span id of the request. Log lines written inside a request get the same ids.
The SDK is in **alpha**. The package is not on a public registry yet. Today it arrives as a tarball that you install with a `file:` dependency. The API is the same on both paths.
## What you get
### Traces
Auto-instrumented spans for the libraries you use, plus [custom spans](/sdk/nodejs/manual-spans) for queue jobs and database calls that have no instrumentation.
### Metrics
`http.server.request.duration` and `http.client.request.duration` for every request, Node.js runtime metrics, the `app.http.server.*` family from the Express or Koa adapter, and your own [custom metrics](/sdk/nodejs/custom-metrics).
### Exemplars
A sampled measurement on every histogram bucket, with the trace id and span id of the request. Your own histograms get them too. The upstream SDK cannot emit exemplars at all ([opentelemetry-js#5147](https://github.com/open-telemetry/opentelemetry-js/issues/5147)).
### Log correlation
The SDK does **not** ship logs. It stamps `trace_id`, `span_id`, and `trace_flags` on every pino log line inside a request, so one filter joins a chart, a trace, and the logs. A pino OTLP transport or a collector sends the lines to Sherlock. See [Ship and correlate logs](/sdk/nodejs/add-telemetry#10-ship-and-correlate-logs).
## Requirements
| Requirement | Value |
| --------------------------------- | ------------------------------------------ |
| Node.js | 20 or later. 20.6 or later for ES modules. |
| Module systems | ES modules and CommonJS |
| TypeScript | Type definitions included |
| Frameworks with a metrics adapter | Express, Koa |
| Peer dependency | `@opentelemetry/api` 1.9.1 |
The package brings its own OpenTelemetry dependencies, pinned to exact versions. Do not add `@opentelemetry/*` packages yourself. Import the trace API from `@sherlock-labs/otel`.
## Next steps
First data in ten minutes.
From credentials to a verified production deploy.
Define your own instruments with exemplars.
Trace queue jobs and database calls.