Honeycomb
by Hound Technology, Inc. (Honeycomb)
OpenTelemetry-native observability that traces AI agents and the systems they touch in one view
Honeycomb is an observability platform built on high-cardinality distributed tracing that now covers AI agents as first-class citizens. Its Agent Timeline renders multi-agent, multi-trace workflows as a single view connecting LLM calls, tool invocations, agent handoffs and downstream system impact, using OpenTelemetry GenAI semantic conventions rather than a proprietary SDK, so engineering teams can debug non-deterministic agent behaviour in production.
Honeycomb, legally Hound Technology, Inc., was founded in 2016 by Christine Yen (CEO) and Charity Majors (CTO) and is headquartered in San Francisco. It has raised approximately $150 million across seed, Series A ($11M, 2019), Series B ($20M, 2020), Series C ($50M, 2021) and Series D ($50M, 2023), and runs a proprietary columnar data store on AWS Graviton ARM infrastructure built for fast, high-cardinality queries over distributed traces. On 12 May 2026 the company launched a set of agent-observability capabilities aimed squarely at the problem that conventional monitoring was not designed for non-deterministic, multi-hop agent workflows. Agent Timeline renders multi-agent, multi-trace workflows as a single view, connecting LLM calls, tool invocations, agent handoffs and system impacts in real time so an agent's decision path can be reconstructed end to end; it shipped in Early Access with general availability anticipated for June 2026. Canvas is a rebuilt collaborative workspace that acts as both a chat interface and an autonomous agent, accepting plain-English queries and producing shareable visualisation snapshots; Canvas Skills encode debugging knowledge into reusable playbooks for frameworks such as Kafka; and auto-investigations trigger when alerts fire, SLOs burn or anomalies surface, gathering data, testing hypotheses and proposing remediation. Critically, the implementation is built on OpenTelemetry GenAI semantic conventions v1.40.0, treating gen_ai.* attributes as first-class so model evaluations, tool executions and MCP calls are captured without re-instrumentation or a proprietary SDK — a deliberate rejection of framework lock-in. Honeycomb holds SOC 2 Type II, is GDPR compliant, signs HIPAA/HITECH BAAs for Pro and Enterprise customers, completes PCI DSS SAQ and CSA STAR Level 1 annually, and offers US and EU data-residency regions.
Platform and SRE leaders already running OpenTelemetry who are now putting agents into production and need the agent's decision path and its downstream blast radius in the same trace store, not in a separate LLM-observability tool.
Reconstruct exactly what a non-deterministic agent did — every LLM call, tool invocation and handoff — and what it changed in production, without re-instrumenting or adopting a vendor SDK.
At a Glance
- Category
- Developer Tools
- Pricing
- Freemium, Usage-based, Contact for pricing
- Target Market
- CTOs, VPs of Engineering, SREs, Platform Engineers, Enterprise Developers
- Deployment
- Cloud-first, Hybrid
- Founded
- 2016
- Headquarters
- San Francisco, United States
Key Features
- ✓Agent Timeline
Renders multi-agent, multi-trace workflows as one view connecting LLM calls, tool invocations, agent handoffs and system impacts, so an agent's full decision path is reconstructable in real time.
- ✓OpenTelemetry GenAI semantic conventions
Implements v1.40.0 with gen_ai.* attributes as first-class citizens, capturing model evaluations, tool executions and MCP calls with no re-instrumentation and no proprietary SDK.
- ✓Canvas
A collaborative workspace that is both a chat interface and an autonomous investigating agent, taking plain-English questions and producing shareable visualisation snapshots for team debugging.
- ✓Canvas Skills
Encodes institutional debugging knowledge into reusable playbooks for frameworks such as Kafka that run autonomously, turning tribal knowledge into repeatable investigation.
- ✓Auto-investigations
Triggers automatically when alerts fire, SLOs burn or anomalies surface, gathering data, testing hypotheses and proposing remediation before an engineer opens the tool.
- ✓High-cardinality columnar store
A proprietary columnar database on AWS Graviton supports fast queries across unbounded dimensions, which is what makes per-request agent debugging feasible at scale.
- ✓Unlimited seats on every tier
Including the free plan, so observability access is not rationed by licence cost and incident response is not blocked on who holds a seat.
Capabilities
Use Cases
- •Agent failure root-cause analysis
Trace a misbehaving production agent through its LLM calls, tool invocations and handoffs to find the step that actually caused the wrong outcome.
- •Agent cost and latency attribution
Correlate gen_ai.* span attributes with request traces to identify which agent steps drive token spend and tail latency.
- •Microservice incident response
Use BubbleUp and high-cardinality queries to isolate which dimension distinguishes failing requests from healthy ones during an outage.
- •SLO-driven reliability management
Define service level objectives, then let auto-investigations gather evidence and propose remediation the moment error budget starts burning.
- •MCP tool-call auditing
Capture and inspect Model Context Protocol calls as first-class telemetry to verify what external tools agents actually invoked in production.
- •Institutionalising debugging expertise
Convert a senior engineer's investigation runbook into a Canvas Skill so on-call engineers reproduce the same analysis without the expertise.
Ideal For
Best For
- ✓Debugging non-deterministic multi-agent workflows where the failing step is not obvious from logs or metrics
- ✓Teams already standardised on OpenTelemetry who want agent telemetry in the same store as application traces
- ✓SRE and platform organisations that need agent behaviour correlated with downstream service impact and SLO burn
- ✓Microservice estates where high-cardinality query performance matters more than dashboard breadth
- ✓Organisations that want agent observability without adopting a framework-specific or proprietary instrumentation SDK
- ✓Engineering orgs that want every developer to have observability access without per-seat licence rationing
Not Ideal For
- ✗Buyers who want a single-vendor full-stack suite covering infrastructure monitoring, log management, RUM and security in one licence — Honeycomb is a tracing-first platform and frontend performance is an Enterprise add-on rather than included
- ✗Teams with unpredictable or very high telemetry volume and no appetite for sampling: event-based metering means cost tracks data volume, and controlling spend usually means running Refinery
- ✗Organisations that need SLOs or a Service Map during a free evaluation — SLOs start at Pro and Service Map is Enterprise-only, so assessing the differentiating features requires a sales conversation
- ✗Teams with no existing OpenTelemetry instrumentation and no capacity to add it, since the platform's value is proportional to trace quality
Integrations
Deployment
Market Analysis
Pros
- ✓Agent Timeline puts multi-agent, multi-trace workflows in a single view with LLM calls, tool invocations, handoffs and system impact connected in real time
- ✓Built on OpenTelemetry GenAI semantic conventions v1.40.0, so adoption requires no re-instrumentation, no proprietary SDK and no framework commitment
- ✓Substantive free tier — 20M events/month, unlimited seats, and Canvas AI Copilot, Honeycomb MCP and Agent Timeline all included
- ✓Ten years of distributed-tracing and high-cardinality query engineering behind it, founded 2016 with roughly $150M raised
- ✓Strong compliance posture for an engineering tool: SOC 2 Type II audited annually, GDPR, HIPAA/HITECH BAAs on Pro and Enterprise, PCI DSS SAQ, CSA STAR Level 1, and US or EU data residency
Cons
- ✗Event-based metering makes cost a direct function of telemetry volume, and practitioners discussing observability spend treat sampling as the necessary control — that is real operational work Refinery exists to do, and Refinery enterprise support is itself an Enterprise-tier item
- ✗The features enterprises evaluate on are gated: the Free tier has no SLOs and no Service Map, Service Map is Enterprise-only, and frontend performance, enterprise alerting and installation services are Enterprise add-ons rather than included
- ✗Not a full-stack suite — it is tracing-first, so most organisations still run something else alongside it for infrastructure monitoring and log management, which undercuts the consolidation argument buyers usually want
- ✗Hacker News commenters have criticised the company's 'observability 2.0 versus 1.0' framing as positioning designed to sidestep direct feature comparison with incumbents rather than a substantive technical distinction
- ✗Agent-specific tooling is materially newer than the tracing core: Agent Timeline shipped in Early Access on 12 May 2026 with general availability only anticipated for June, so it has a short production track record relative to the platform
- ✗Honeycomb is not itself ISO/IEC 27001 certified — its documentation describes services delivered through certified AWS and GCP environments, which is a weaker claim than a direct certification
Pricing
Free
$0
- ✓Up to 20M events/month
- ✓Up to 100M metric data points/month
- ✓Unlimited seats
- ✓2 Triggers
- ✓Distributed tracing and BubbleUp
- ✓OpenTelemetry support
- ✓Canvas AI Copilot, Honeycomb MCP and Agent Timeline
- ✓No SLOs, no Service Map
Pro
From $150/mo
- ✓Up to 750M events/month
- ✓Up to 3.75B metric data points/month
- ✓Unlimited seats
- ✓100 Triggers
- ✓2 SLOs
- ✓Single Sign-On (SSO)
- ✓Honeycomb Support
- ✓Monthly or annual billing
Enterprise
Contact for pricing
- ✓Base of 10B events/year, variable beyond
- ✓300+ Triggers and 100+ SLOs
- ✓Service Map
- ✓Query Data API and SLO Reporting API
- ✓AWS PrivateLink
- ✓Private Cloud support
- ✓Enterprise support for Refinery
- ✓Onboarding and installation services
Telemetry pipeline add-on
From $0.10/GB
- ✓Telemetry collection billed per gigabyte
- ✓Available alongside any tier
Billing is metered on events per month rather than hosts or seats, and seats are unlimited on every tier including Free, so cost scales with telemetry volume and not headcount. Free covers 20M events and 100M metric data points monthly; Pro starts at $150/month for up to 750M events and 3.75B metric data points and is the first tier with SSO and SLOs; Enterprise is quote-only, starting from a 10B events/year base and adding Service Map, AWS PrivateLink, Private Cloud, the Query Data and SLO Reporting APIs and enterprise Refinery support. Telemetry collection is a separate add-on from $0.10/GB, and frontend performance, enterprise alerting and installation services are Enterprise-tier extras. Budget realistically for sampling via Refinery, since uncontrolled high-cardinality volume is the main driver of overspend.
Security & Compliance
Connect
Sources
This page was written from 6 sources, 4 on domains other than honeycomb.io.
- 1.honeycomb.io — pricingvendor
- 2.honeycomb.io — honeycomb launches agent observability full visibility agentvendor
- 3.docs.honeycomb.io — compliance data privacy
- 4.prnewswire.com — honeycomb launches agent observability bringing full visibil
- 5.en.wikipedia.org — Honeycomb (company)
- 6.hn.algolia.com — hn.algolia.com
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Tricentis Agentic Quality Engineering Platform
A governed fleet of testing agents that validates AI-written enterprise software before it ships
AlphaEvolve
Gemini-powered evolutionary agent that discovers faster algorithms for your hardest optimization problems
Greptile
AI code review agent that indexes the whole codebase to catch cross-file bugs before merge
GitLens
The Git workbench where human and AI-agent changes get reviewed, composed and merged in one view