Skip to main content
Sherlock stores the logs, traces, and metrics your services send over OpenTelemetry and links them by trace id. A spike on a chart opens the request that caused it. The request opens its trace. The trace opens its logs.
Prefer to ask instead of click? Connect Claude Code, Cursor, or another MCP client to Sherlock and ask “How many errors did we get in the last hour?” or “Latency spiked at 14:20. Check the metrics, then pull the traces behind it.” See Ask an AI assistant.Working with a coding agent? Give it these docs through the documentation MCP server. See Docs for AI agents.

​
How Sherlock works

Three things happen, in order:
  • Send. Your services export OTLP over HTTP to the Sherlock ingest endpoint with a bearer token. An env resource attribute routes each service’s data into a source.
  • Explore. The Logs, Traces, and Metrics pages search and chart each signal on its own.
  • Pivot. trace_id joins the three. An exemplar on a histogram carries the trace id of one request, so a chart leads to a trace and a trace leads to its log lines.
Using Node.js? The Sherlock SDK for Node.js is the only Node.js distribution that puts trace-linked exemplars on histograms. Start with the Node.js quickstart.

​
Ways to send data

​
Node.js SDK

The Sherlock distribution of OpenTelemetry for Node.js instruments a service with one import, sends traces and metrics, and adds exemplars to every histogram. Your logs keep their own pipeline, stamped with the trace id.

​
Go

The upstream OpenTelemetry Go SDK works with Sherlock as it is. One bootstrap file turns on delta metrics, exemplars with trace flags, and the seconds histogram boundaries. Traces and metrics come from otelhttp, and a slog handler stamps the trace id on every log line.

​
Other OpenTelemetry SDKs

Any OpenTelemetry SDK with an OTLP/HTTP exporter can send to Sherlock. Point the exporter at the ingest endpoint, add the bearer token, and set the env attribute.

​
OpenTelemetry Collector

An OpenTelemetry Collector fans in many services and forwards logs, traces, and metrics to Sherlock through one exporter. This is the usual path for logs.

​
Where to explore

  • Logs. Search, filter by field, open the trace behind a line.
  • Traces. Find a trace, read the waterfall, jump to its logs.
  • Metrics. Chart a metric, group by labels, click an exemplar.
  • Exemplars. Follow a spike to the request and its logs.
  • Alerts. Alert rules on logs and metrics, with notification channels.

​
Where you can use Sherlock

  • Web app. The Logs, Traces, Metrics, Alerts, and Settings pages.
  • AI assistants. The MCP server exposes the same logs, traces, and metrics to Claude Code, the Claude apps, Cursor, VS Code, Codex, and any other MCP client.

​
Next steps

Get started

Create an organization and send your first data.

Node.js quickstart

First traces and metrics from a Node.js service in ten minutes.

Go setup

The upstream Go SDK with one bootstrap file.

Ask an AI assistant

Connect an AI client through MCP.

Docs for AI agents

Connect an agent to the documentation MCP server.

Explore your data

Logs, traces, metrics, exemplars, and alerts.