Along with Agent Timeline, Canvas Agent and Expertise supercharge engineering groups and their AI brokers with a shared manufacturing observability layer
SAN FRANCISCO, Might 12, 2026 /PRNewswire/ — Honeycomb.io, the observability platform constructed for the brand new form of software program, at this time launched a sequence of agentic intelligence and agent observability options purpose-built for AI brokers in manufacturing. These new and enhanced capabilities embody Agent Timeline, Canvas Agent, and Canvas Expertise. Engineering groups will now have real-time visibility into what their brokers are literally doing with out proprietary software program improvement kits (SDKs) or framework lock-in.
AI brokers are quickly dealing with duties reminiscent of code era, incident triage, cloud infrastructure deployment, and customer support. Nonetheless, the instruments constructed to watch software program methods weren’t designed for non-deterministic, multi-hop agent workflows. Dashboards break down, averages lie, and when an agent causes an incident, groups don’t have any approach to reconstruct what it determined or why.
“AI has upended how software program works: introducing actual nondeterminism into manufacturing, and essentially altering how groups of people and brokers construct and validate code,” mentioned Christine Yen, cofounder and CEO of Honeycomb. “Engineers are drowning in uncertainty as most observability instruments weren’t constructed for this form of “unknown unknown.” Honeycomb was constructed for the toughest elements of constructing software program—and autonomous brokers have taken these hardest elements mainstream.”
Observability for AI methods
Honeycomb’s new AI-native observability capabilities permit Engineering groups to trace and analyze the efficiency, habits, and interactions of AI brokers. With Agent Timeline, customers can render multi-agent, multi-trace workflows as a single coherent view, connecting each LLM name, device invocation, agent handoff, and downstream system impression in actual time. Engineering groups can now hint what an agent did, reconstruct the total determination path, and perceive failures with out switching instruments or piecing collectively logs by hand.
Observability with AI brokers
New and enhanced options from Honeycomb supercharge human and agent groups by surfacing intelligence at machine pace. These embody:
- Canvas, newly rebuilt to function a collaborative workspace, a chat interface, and autonomous agent multi function, permits engineering groups to question points in plain English, work with human and agent workforce members on investigations, and produce sharable visualization snapshots.
- Auto-investigations set your Canvas agent to work robotically when an alert fires, an SLO burns, or an anomaly surfaces, gathering information, creating and testing hypotheses, and proposing remediation, all earlier than an engineer even opens their laptop computer.
- Canvas Expertise encode your finest engineers’ debugging data and finest practices for frameworks or providers like Kafka into reusable playbooks that run autonomously.
“Honeycomb is nice at coping with unknown-unknowns, and Canvas takes it even additional,” mentioned Shogo Wada, Employees Software program Engineer at Bubble. “Throughout a latest investigation, we discovered a reason for slowness in our API by asking Canvas to research. Whereas the trigger was in a roundabout way seen on the affected spans, Canvas in contrast entire traces and located patterns inside their little one spans. Earlier than Canvas, this could have been a handbook means of opening traces one after the other. Moreover, as Canvas is a collaborative device among the many workforce, we may theoretically discover a number of hypotheses in parallel, attending to a root trigger quicker.”
Constructed on open requirements for max flexibility
As a distinguished supporter and contributor to the OpenTelemetry mission, Honeycomb has built-in the OpenTelemetry (OTel) GenAI semantic conventions (v1.40.0) into its platform. By making gen_ai.* attributes first-class residents, Honeycomb ensures that mannequin evaluations, device executions, MCP calls, LLMs, and brokers are all correctly noticed. This alignment with OpenTelemetry GenAI requirements permits for the automated era of structured GenAI insights with out requiring re-instrumentation as specs change, customized setups, or the usage of proprietary SDKs.
Honeycomb’s proprietary unified information retailer was constructed for precisely the sort of unpredictable, high-dimensionality telemetry that AI brokers generate, from the bottom up.
Availability
Canvas, Canvas Agent, and Expertise can be found beginning subsequent week for all Honeycomb prospects. Agent Timeline is accessible in Early Entry, and anticipated to be usually out there subsequent month.
Expertise Honeycomb first-hand
Honeycomb is internet hosting a three-day digital occasion, Innovation Week: Observability for the Agent Era, to showcase these new capabilities. The digital occasion runs Might 12-14 and can function product deep dives, stay demos, and buyer periods. Be a part of us stay or on-demand.
O11yCon, Honeycomb’s multi-day flagship convention, follows on Might 20-21 in San Francisco. The occasion brings collectively builders, SREs, platform groups, and engineering leaders for 2 days of workshops, practitioner periods, and a management monitor centered on working AI-powered methods in manufacturing. Registration is open now.
Assets
About Honeycomb
Honeycomb is the observability platform constructed for the brand new form of software program. Constructed on a decade of distributed tracing management and deep roots within the OpenTelemetry group, Honeycomb offers engineering groups—and the AI brokers now working alongside them—real-time, high-cardinality solutions about any manufacturing system, with no pre-aggregation and no cardinality limits. Study extra at www.honeycomb.io and observe us on LinkedIn.
Media Contact
Ciri Haugh
[email protected]
SOURCE Honeycomb
