Honeycomb Observability Platform

Honeycomb is an observability platform that helps engineering teams understand how their code behaves in production by collecting high-cardinality event data from distributed services. Built on an OpenTelemetry-native architecture, the platform ingests traces, metrics, and logs as structured events rather than pre-aggregated time series, allowing engineers to slice and group data by any dimension at query time without losing context. The BubbleUp feature automatically detects anomalies by comparing outlier events against the baseline distribution, highlighting fields that correlate with the issue and reducing mean-time-to-resolution from hours to minutes. Queries return results in under ten seconds across billions of events, with support for aggregations, group-by operations, time comparisons, and mathematical functions. The platform has added specific capabilities for AI and LLM observability, capturing token usage, model latency, prompt and response data, and agent execution traces so teams can debug non-deterministic AI workflows the same way they debug traditional distributed systems. An AI copilot called Canvas enables natural language investigation of production data, while an MCP integration lets coding agents query telemetry directly during development. Service Level Objectives track error budgets and burn rates, triggering alerts only when reliability targets are at risk. Telemetry pipelines provide data filtering, sampling, and routing controls to manage observability costs. The platform supports unlimited cardinality fields and unlimited team members at no additional per-seat cost, with private cloud deployment options for organizations requiring data residency in their own AWS accounts. Customers include Slack, Dropbox, Intercom, HelloFresh, and Booking.