Honeycomb

What Honeycomb does

Honeycomb is an observability platform built for the new shape of software, including AI agents. It provides engineering teams with real-time, high-cardinality answers about any production system, with no pre-aggregation and no cardinality limits. Built on a decade of distributed tracing leadership and deep roots in the OpenTelemetry community, Honeycomb helps engineering teams follow their code into production, from tracing distributed services to debugging non-deterministic AI workflows. The platform offers complete visibility with data available in under 90 seconds, allowing engineers and AI agents to solve problems quickly. It enables sub-10 second queries, root cause analysis in under three minutes with BubbleUp, and infinite scalability without making tradeoffs between observability cost and software quality. Honeycomb supports unlimited fields and users at no extra cost, with no vendor lock-in, and is designed to be OpenTelemetry-native. It helps debug LLM applications with confidence, providing granular insight into how agents and LLMs behave in production, troubleshooting failures faster, and continuously improving model performance in real-time with real data. Honeycomb's unified data model retains high-value relationships across production data, and its purpose-built columnar data store delivers sub-second query times, even across high-cardinality data. The platform also offers predictable pricing, allowing teams to control their observability budget.

Products

• Honeycomb Free PlanIntroductory plan, free forever, best for testing and individual projects. Includes 2 Triggers, Distributed Tracing, BubbleUp, OpenTelemetry Support, Team Query History, Query Result Permalinks, Honeycomb Metrics, Canvas AI Copilot, and Honeycomb MCP. Event Volume: Up to 20M per month. Metrics Data Points: Up to 100M per month.
• Honeycomb Pro PlanBest for teams with a production application. Includes all features from the Free Plan, plus 100 Triggers, 2 Service Level Objectives (SLOs), Single-Sign On (SSO), Honeycomb Support, and Monthly or Annual Subscription. Event Volume: Up to 750M per month. Metrics Data Points: Up to 3.75B per month. Starting at $150 per month.
• Honeycomb Enterprise PlanBest for multi-team and large-scale applications. Includes all features from the Pro Plan, starts with 300 Triggers, starts with 100 SLOs, Service Map, Enterprise Support for Refinery, Enterprise-Grade Support + Onboarding, Query Data API, AWS PrivateLink, SLO Reporting API, and Support for Honeycomb Private Cloud. Custom plans to fit needs. Event Volume: Variable. Metrics Data Points: Variable. Starts with a base allowance of 10 billion events per year.
• Honeycomb CanvasAI-assisted copilot for investigations, allowing natural language queries and interactive data exploration. Enables any developer or SRE to debug like an expert with AI-assisted investigations.
• Honeycomb MCP (Model Context Protocol)Connects AI agents directly to observability data, allowing analysis, visualization, and reasoning about telemetry data through a secure, standards-based interface. Access observability data directly from an IDE.
• Honeycomb MetricsMetrics and charts with high-cardinality data, making metrics debuggable. Uses a time series database for infrastructure metrics and an events-based model for application and business metrics with unlimited cardinality.
• Honeycomb TracesFind interesting traces with uniquely powerful trace analytics, waterfall views to identify slow or failing components. Debug massive LLM traces with thousands of tokens as easily as traditional API calls.
• Honeycomb EventsStructured log events as the building block of the platform. Easily derive logs, metrics, and traces from these events, with a powerful query engine for fast analysis.
• Honeycomb IntelligenceBrings AI into investigations, supercharging them with AI-assisted workflows.
• Honeycomb Anomaly DetectionMachine learning-powered tool to surface anomalies in any data (metrics, traces, logs).
• Honeycomb Agent TimelineCaptures GenAI Conversations per request or agent run, with child spans for LLM calls, tool invocations, and handoffs. Provides visibility into agent decisions, tool calls, handoffs, and token cost across every production workflow.
• Honeycomb Query AssistantAI-powered assistant to ask questions of the system in plain English, eliminating the need for complex query languages.
• Honeycomb Telemetry PipelineHelps enterprises get more value out of their telemetry, enabling collection, transformation, and management of data with ease.
• Honeycomb Private CloudAllows organizations to bring Honeycomb's observability platform into their own private AWS environment, maintaining control over data, compliance, and infrastructure. Includes AI-native tooling like Canvas, MCP, and anomaly detection powered by AWS Bedrock.
• Honeycomb Frontend ObservabilityUnifies frontend data and allows quick querying from frontend to backend for faster troubleshooting, going beyond real user monitoring (RUM).
• Honeycomb Log AnalyticsComb through billions of log lines in seconds, spot anomalies quickly, and resolve issues with precision. Unifies logs, traces, and metrics in one tool.

Key capabilities

• AI-Ready Observability
• Real-time, high-cardinality answers
• No pre-aggregation and no cardinality limits
• Distributed Tracing
• OpenTelemetry-native platform
• Sub-10 second queries
• Root cause analysis with BubbleUp
• Unlimited fields and users
• Predictable pricing model
• Granular insight into LLM behavior
• Troubleshoot failures faster
• Continuously improve model performance
• Unified data model
• Purpose-built columnar data store
• Natural language querying with Canvas
• IDE integration with Honeycomb MCP
• AI-assisted investigations
• Metrics with high-cardinality data
• Trace analytics with waterfall view
• Structured log events
• Anomaly Detection
• Agent Timeline for AI agents
• Query Assistant
• Service Level Objectives (SLOs)
• Service Map
• Frontend performance analysis add-on
• Enterprise-grade alerting add-on
• Installation and enablement services add-on
• Query Data API
• AWS PrivateLink
• SLO Reporting API
• Support for Honeycomb Private Cloud
• Agent-ready observability for custom metrics
• Free custom metrics
• Modern approach for high cardinality data
• Telemetry Pipeline for data strategy

Use cases

• Following code into production
• Tracing distributed services
• Debugging non-deterministic AI workflows
• Real-time understanding of end-user experience
• Troubleshooting LLM failures
• Improving LLM model performance
• Investigating and debugging production issues before customer impact
• Optimizing observability cost and software quality
• SLO-based monitoring for LLM reliability
• Catching and investigating anomalies without noisy alerts
• AI-assisted debugging for developers and SREs
• Debugging like an expert with AI-assisted investigations
• Collaborative debugging
• Analyzing metrics with high-cardinality data
• Identifying slow or failing components in traces
• Analyzing structured log events
• Supercharging investigations with AI
• Solving unpredictable problems in AI systems (nondeterministic behavior, expensive LLM calls, multi-step agent workflows)
• Understanding AI system behavior with Agent Timeline
• Pinpointing issues and conducting investigations
• Detecting anomalies and identifying solutions
• Tracing agent decisions, tool calls, handoffs, and token cost across production workflows
• Controlling cost as AI agents are shipped
• Catching agents going off the rails (loops, retries, runaway agents)
• Debugging high-cardinality agent behavior
• Reducing observability costs
• Accelerating adoption of observability across teams
• De-risking migration and tool retirement
• Improving incident resolution times
• Reducing unplanned outages
• Increasing DevOps team productivity
• Discovering issues in AWS infrastructure
• De-risking AWS applications
• Optimizing performance of AWS services
• Migrating live architecture components seamlessly
• Streamlining log analysis
• Debugging issues in minutes with log analysis
• Implementing a telemetry data strategy
• Collaborating across teams during incidents
• Maximizing log value while minimizing costs
• Finding issues before customers do
• Fixing frontend issues fast
• Debugging visually with Service Map
• Monitoring and debugging together as a team
• Setting error budgets aligned with business goals

AI approach

Honeycomb provides an AI-ready observability platform that helps engineering teams monitor, debug, and understand the behavior of AI agents and LLMs in production. It offers AI-assisted investigation tools like Canvas (AI copilot) and Honeycomb MCP (Model Context Protocol) for natural language querying and direct IDE access to observability data. The platform focuses on providing granular insight into non-deterministic AI workflows, tracing agent decisions, tool calls, handoffs, and token costs. It also offers anomaly detection and query assistance using AI.

Tech named: LLMs, AI agents, Canvas AI Copilot, Honeycomb MCP, BubbleUp, Query Assistant, Anomaly Detection, Claude Code, Amazon Bedrock AgentCore

Industries served

• Software Development
• Gaming
• Financial Services
• E-commerce
• Education
• Banking

What it says sets it apart

• Observability built for the AI era, supporting AI agents and LLMs
• Real-time, high-cardinality answers with no pre-aggregation and no cardinality limits
• OpenTelemetry-native platform with no vendor lock-in
• Sub-10 second queries and root cause analysis in under three minutes with BubbleUp
• Predictable pricing model that scales with application complexity, not penalizing curiosity or increased data volume
• Unified data model that retains high-value relationships across production data (logs, metrics, traces)
• Purpose-built columnar data store for sub-second query times on high-cardinality data
• AI-assisted investigations with Canvas and IDE integration with Honeycomb MCP
• Agent Timeline for detailed visibility into AI agent behavior, including decisions, tool calls, handoffs, and token cost
• Ability to debug massive LLM traces with thousands of tokens easily
• Free custom metrics and unlimited metadata enrichment without extra cost
• Focus on solving 'unknown unknowns' that traditional monitoring approaches don't address
• Superior for LLM reliability compared to traditional metrics, especially with SLO-based monitoring
• Seamless onboarding and integration with existing infrastructure using OpenTelemetry standards
• Significant cost savings and improved user experience compared to consolidating multiple tools
• Ability to trace every agent run end to end, including cost per request, feature, and outcome
• Visibility into anomalous tokens per session, repeated tool calls, and runaway agents
• Debugging high-cardinality agent behavior at the individual customer or session level
• Offers Honeycomb Private Cloud for full control over data, compliance, and infrastructure in a private AWS environment