Honeycomb

Honeycomb, made by Hound Technology, Inc., is an observability platform that gives engineering teams and AI agents a shared view of how software behaves in production. It combines distributed tracing, log analytics, metrics, SLOs, and AI-focused tooling built on structured, high-cardinality event data.

What It Is

Honeycomb is a production observability platform. Engineers send telemetry (logs, metrics, traces) and query it through a columnar data store; the site describes sub-second queries across high-cardinality data. Data can be explored as traces in a waterfall view, as structured log events, or through visualizations such as heatmaps. BubbleUp detects outliers and helps explain what differs in affected events, and SLOs let teams define acceptable service levels and alert on them.

AI-assisted investigation and AI workload observability

Canvas lets users query in natural language, run automated investigations in a shared workspace, and the site says it runs investigations when SLOs burn. The Honeycomb MCP server gives coding agents and IDEs access to production observability data, and Agent Skills encode engineers' playbooks. For teams running AI systems, Agent Timeline shows LLM calls, tool invocations, agent handoffs, and downstream spans in order, with token cost and latency tracked per model. Anomaly Detection is also listed.

Data collection and OpenTelemetry

Honeycomb describes itself as OpenTelemetry-native, with support for instrumenting many languages. Related platform pieces include a Telemetry Pipeline, Refinery as a Service for sampling, frontend observability, and a Private Cloud option. The company states it is SOC 2 Type II certified.

Getting started and support

Users can sign up for a free account, request a demo, or try a sandbox that requires no signup. Documentation, an academy, community Slack, office hours, and email and in-app support are available.