Improvement Canvas Sections Canvas now groups each line of investigation into a section. A section has a summary card with the question, what the agent found, the evidence, a conclusion, and next steps. The charts and tables it cites sit next to it, in the order the summary references them to make following the agent's investigation easier. Clicking on any piece of evidence will navigate you direc…
Improvement When you connect an AI assistant to the Honeycomb MCP, you no longer have to pick between a blanket read-only and full read-write access. You can now choose access for each resource. For example, you can let your agent create boards but only read SLOs and triggers. On the consent screen, choose one of: Read all: your client can read every resource. Read and write all: your client can r…
New Canvas now remembers important context across all your investigations. Canvas saves what it learns about your systems, and you can ask it to remember something. Canvas uses those facts in your next investigation, so you explain less. Your memories are personal. Only you see them. Open Canvas > Memory to see, add, edit, or delete them. You can also ask Canvas to forget something in chat. If you…
New Quote Text in Canvas Chat Need to ask the Canvas agent about its findings? Select the text and click Ask in chat. Your question is attached as context so you don't have to describe what you mean.
Integrations New The Canvas Agent can now connect to GitLab, LaunchDarkly, Stripe, CircleCI, Intercom, and Amp. GitLab: read merge requests, commits, pipelines, job logs, and issues. With your approval, comment, open merge requests, or retry pipelines. LaunchDarkly: read feature flags and targeting rules. With your approval, create or update them. Stripe: read payments, subscriptions, invoices, an…
New Once data starts flowing into Honeycomb, you don't have to wonder what to do next. In the Data Pulse experience after you send data, Honeycomb reads your incoming telemetry to work out what kind of system you’re running, then points you to the questions worth answering first. A read on your system: Data Pulse looks at your data and summarizes what your environment appears to be, with a breakdo…
New Sending your data to Honeycomb is now easier. A short wizard points you to the setup that fits you, so you spend less time figuring out where to start and more time seeing what your systems are doing. Choose your pace: Quickstart takes you straight to an ingest key and instructions. The managed experience walks you through it. Choose how to connect: Get started with the MCP, our recommended ro…
Integrations Canvas now exclusively uses the Honeycomb GitHub App to connect and authorize connections to GitHub. The Canvas GitHub connector lets Canvas read your code so it can better identify issues, improve telemetry, and diagnose alerts. Previously, Canvas required both the installation of the Honeycomb GitHub App and individual users to authenticate. Now the connection is made once per team.…
New The Canvas agent now includes direct links to the shapes it generates. This enables you to jump directly to and between charts that are referenced. Click into any of the shape links provided to see the information cited!
We've raised the minimum password length to 16 characters. This applies only when you create an account or reset your password. Existing passwords aren't affected.
New Agents can now create and update SLO Burn Alerts through the Honeycomb MCP. They can also reset SLO budgets. These tools allow agents to create full observability-driven monitoring and alerting loops with Honeycomb SLOs. To access the new tools, connect to Honeycomb MCP with Read and Write permission. Learn more about Honeycomb MCP and Honeycomb SLOs.
Improvement By default, a grouped trigger sends no resolution notification until every group clears, and uses the same set of recipients regardless of which group fired. Group resolution alerts is a new third option under Frequency of Alerts. Select it on a grouped trigger to get: Per-group resolution. Each group resolves independently when it drops below threshold, and the notification names the…
Improvement We've refreshed the visual design across the Honeycomb UI. The result is a more consistent, readable interface that gets out of your way while you investigate. What you'll notice: Rounder, softer UI. Inputs, buttons, cards, menus, and modals are easier to tell apart from the page and from each other, so you can scan a screen and find the control you need faster. Canvas is easier to rea…
New Canvas Connectors are in public beta. Canvas Agents already connect to GitHub and Linear. They now also connect to Amplitude, Atlassian, Braintrust, Checkly, incident.io, Notion, Pylon, and Sentry. Connectors give the Canvas Agent context from outside Honeycomb. For example, the Canvas Agent can combine the failing Checkly check, the Sentry stack trace, the Github code change, and the Jira iss…
New Get board suggestions tailored to your stack. Home now surfaces board templates matched to the data you have — no more digging through generic templates to find one that fits. When Honeycomb detects data from a recognized service or stack, you'll see a "For your stack" suggestion on the Home page with a ready-to-use board template, like performance tracking boards built for your services. Clic…
New New to Honeycomb? You no longer need to instrument your own app to see what Honeycomb can do. The Send Data page now includes an Explore with Sample Data option. Pick a scenario that matches what you're building, and we'll send a realistic trace straight into your environment — no code, no config, no waiting on your own instrumentation. Choose from three scenarios: E-commerce: a web checkout f…
Improvement Your Canvas Agents can now observe and warn you about patterns and pin-point repeat issues faster in your auto-investigated Triggers, Burn Alerts, and Anomalies. Honeycomb now dispatches Automatic Investigations to Canvas agents with awareness of recent firings of those alerts. Canvas agents maintain information about your alerts over time, letting them build on previous hypotheses and…
Improvement You can now search for any span name in a conversation - tool calls, LLM operations, agent invocations, and more. Matching spans are highlighted in the timeline so you can jump straight to the steps you care about instead of scanning a long-running conversation end to end.
New Getting your data into Honeycomb just got a lot faster. Connect your codebase directly through the Honeycomb MCP server, and get guided step-by-step from zero to insight — right from wherever you're already working with AI. With MCP-based onboarding, you can: Send data in minutes: Install the Honeycomb plugin, connect via MCP, and auto-instrument your app with a guided prompt that walks you th…
New Anomaly Detection is now in open beta, and it's on for your team today. Anomaly Detection establishes a statistical baseline of normal behavior for each of your services, then notifies you when something deviates from it. No thresholds to define, no queries to write, nothing to configure. Honeycomb identifies which of your services have enough steady traffic to model reliably, onboards them au…
New The Activity Log environment now includes the telemetry_stats dataset. This dataset records the number of events sent to Honeycomb per environment, dataset, API key, and event category, as well as the outcome (whether we accepted or rejected the event, or encountered an error), and any error message. There's also a new Telemetry Stats board template, with a variety of useful queries for monito…
New You can now create or update Triggers, SLOs, and Boards using Canvas. Edits always require human approval and are not available to the agent during automatic investigations.
New Connectors allow you to bring your tools into the canvas. Available connectors in beta: Linear Github
New When authorizing an MCP client to the Honeycomb MCP Server, you may now select multiple teams at once. Each team can have unique read/write permissions. If multiple teams are selected, your agent will need to pass a team parameter when making tool calls to disambiguate.
Improvement Filter by any custom attribute - email, user ID, agent name, and more to land on the Conversations that matter. It's the fastest path from "here are all of my Conversations" to "here's the one Agent Conversation I need to debug."
New Activity Log has graduated from beta and is now generally available for Honeycomb Enterprise customers. Activity Log gives you a detailed record of user activity and resource changes in Honeycomb — who changed what, and when. Since launching in open beta, teams have been using it for: Change investigation: Trace account changes, or configure alerts on critical updates via - Triggers on Activit…
Improvement When configuring who receives usage notifications, you can now send them to any email address — not just Honeycomb team members. This opens up broader integration possibilities, such as: Routing usage notifications to a Slack channel via a channel email address Sending notifications to a shared team alias or distribution list Forwarding usage alerts into other platforms that accept inb…
New Agent Timeline gives you visibility into your multi-agent, multi-trace AI workflows - LLM calls, tool invocations, token, messages, and failures, all in one place. Here's what's included: Conversations View: browse agent conversations across the last 60 days with agents, tool calls, tokens, P95 latency, and failure counts at a glance Agent Timeline: open any conversation to see duration, model…
New This release brings a redesigned investigation surface built for how teams work today: with multiple humans and multiple agents collaborating on the same investigation in real time, agents that act autonomously when alerts fire, executable knowledge that teaches them how your systems work, and connections to the tools where your code lives. Auto-investigations When a trigger fires, an SLO burn…
New BubbleUp Insights analyzes your BubbleUp results for you to quickly surface notable differences and potential root causes in your data. Insights will load automatically when running a BubbleUp in Query Results for all teams opted in to the Honeycomb Intelligence Terms. You can still analyze and interact with BubbleUp yourself using the charts below the Notable Differences section.