Grafana Labs blog on Grafana Labs

Grafana 13.3 release: manage variables across dashboards, share findings with Notebooks, a UI refresh, and more

Grafana Labs Team — Thu, 29 Oct 2026 15:01:42 +0000

Grafana 13.3 is here, bringing more consistency and control to how you work across Grafana.

Cross-dashboard variables help you manage shared dashboard controls from one place, while Notebooks let you capture and share findings from across Grafana in a single, editable document. You'll also get a first look at Grafana's biggest visual refresh since Grafana 7, along with new ways to guide and refine dashboards built with Grafana Assistant.

Below are just some of the highlights from this release. You can explore all of the updates in the changelog or our What's new in Grafana 13.3 documentation.

Define a variable once and use it across dashboards

Available in public preview in all editions of Grafana

When multiple dashboards use the same variable, such as an environment, region, or cluster selector, managing it separately in each dashboard creates duplicate work and can lead to configuration drift. To address this, we’re rolling out cross-dashboard variables, which let you define a variable once and reuse it across dashboards in your Grafana organization.

You can create and manage these variables from Dashboards > Variables, and scope each one as either:

Global: available to every dashboard in the organization
Folder: available only to dashboards in the same folder

When you edit a dashboard, add a variable from the sidebar and select Global or folder variable. Grafana shows only the variables available to that dashboard's scope, and you can add them one at a time or all at once. Viewers can still change the variable's value on the dashboard, while its configuration stays managed centrally, so shared controls stay consistent without editing every dashboard that uses them.

To try cross-dashboard variables in self-managed Grafana, enable the grafana.dashboardGlobalVariables feature toggle. To learn more, refer to the cross-dashboard variables documentation.

Consolidate and share your findings with Notebooks

Available in public preview in all editions of Grafana

An investigation rarely happens in one place. You might start on a dashboard panel, jump into Explore to run a query, and then ask Grafana Assistant what's going on. Until now, the evidence and your reasoning ended up scattered across tabs, screenshots, and chat threads.

Notebooks let you capture findings from across Grafana in a single, editable document. Add panels from dashboards or Explore, pull in findings from Grafana Assistant, and write your observations right next to the data. Teammates can then follow your investigation, understand how you reached your conclusions, and collaborate on next steps.

You can use a notebook to:

Explore a question about your data, keeping queries and notes side by side as your thinking develops.
Document an investigation so the next person doesn't have to start from scratch.
Prepare a post-incident review with the relevant panels already in place.

You can create notebooks manually or with Grafana Assistant, then save and share them for future reference. The new Notebooks page also makes it easy to find and manage your work, with options to search, filter, edit, and export notebooks.

Notebooks are gradually rolling out to Grafana Cloud. To try them in self-managed Grafana, enable the dashboard.notebooks feature toggle. To learn more, refer to the Notebooks documentation.

A fresh new look for the Grafana UI

Available in public preview in all editions of Grafana

Whether it’s adding new data sources or supporting new use cases, we’re constantly expanding what Grafana can do. And as AI-powered workflows with Grafana Assistant became a bigger part of how people use Grafana every day, we saw an opportunity to rethink how the platform looks and feels.

With Grafana 13.3, we're introducing the first phase of our biggest visual refresh since Grafana 7. Built on an updated set of design principles, the refresh introduces new colors, typography, and UI components to create a more modern, consistent experience across Grafana.

Updates include:

New default dark and light themes. We've reworked every theme color, with more shades for flexibility, and main action buttons now have a solid, high-contrast style.
Softer borders and floating components. Rounder corners offer a more modern feel, and subtle borders on cards, panels, and drawers create a clearer sense of depth. The toolbar and navigation also stand apart more clearly from page content.
Updated inputs, drop-downs, buttons, and tags. We've resized our most-used components to be more proportional and consistent.
Revised typography. A new type scale mixes font weights and sizes, rather than relying on size alone, so it's easier to see what matters most on a page.

Importantly, this is a visual refresh, not a redesign. Your existing workflows and navigation stay the same, and there are no breaking changes for plugins. Plugins that use standard Grafana components and theme tokens will automatically pick up the new look, while those with custom styling can migrate on their own timeline.

To try out the new UI in self-managed Grafana, enable the grafana.visualDesignRefresh feature toggle. You can switch back to the previous design at any time, and we welcome your feedback as we continue rolling out improvements over the coming months.

Build and edit dashboards with more control using Grafana Assistant

Grafana Assistant can already help you build and edit dashboards from a plain-language description. Now, with Grafana 13.3, you have more control over that process, from reviewing and refining a dashboard plan before it's built to easily undoing changes afterward.

Review a dashboard plan from Assistant

Available in public preview in all editions of Grafana

With dashboard plans, you can describe what you want to monitor, and instead of building it right away, Assistant proposes a plan in chat. The plan covers the dashboard title, tabs or rows, panels, visualization types, and variables, so you can catch a missing panel or the wrong chart type before anything is built.

Select Preview to open a read-only version of the dashboard, then keep the conversation going to iterate from there. When the preview matches what you need, select Build it to generate the dashboard and save it.

Quickly revert Assistant edits

Generally available in all editions of Grafana

Experimenting with different layouts and visualizations is easier when you can quickly undo changes you don’t want. With Grafana 13.3, you can now select Undo turn after Grafana Assistant edits your dashboard to restore it to its previous state. If Assistant changed multiple panels in a single request, you can undo all those changes at once.

To learn more, refer to the Grafana Assistant dashboarding guide.

More to explore in Grafana 13.3

Beyond these highlights, Grafana 13.3 includes a range of updates related to data visualization, alerting, Git Sync, and more.

For an in-depth list of all the new features, check out our Grafana documentation, the Grafana changelog, or our What's New documentation.

Join the Grafana Labs community

We invite you to engage with the Grafana Labs community forums. Share your experiences with the new features, discuss best practices, and explore creative ways to integrate these updates into your workflows. Your insights and use cases are invaluable in enriching the Grafana ecosystem.

Upgrade to Grafana 13.3

Download Grafana 13.3 today or experience all the latest features by signing up for Grafana Cloud, which offers an actually useful forever-free tier and plans for every use case. Sign up for a free Grafana Cloud account today.

Our Grafana upgrade guide also provides step-by-step instructions for those looking to upgrade from an earlier version to ensure a smooth transition.

Special thanks to our community

We extend our heartfelt gratitude to the Grafana community!

Your contributions, ranging from pull requests to valuable feedback, are crucial in continually enhancing Grafana. And your enthusiasm and dedication inspire us at Grafana Labs to persistently innovate and elevate the Grafana platform.

Evaluate AI models for observability: o11y-bench joins Fireworks' Specialized Intelligence Index and reaches v1.1

Sean Porter — Wed, 28 Oct 2026 16:08:24 +0000

When evaluating AI models for observability, generating a plausible query or explanation isn't enough. Teams need to know whether an agent's conclusions are supported by telemetry, whether its changes to dashboards actually work, and whether it can deliver reliable results consistently.

This is why we built o11y-bench, Grafana Labs' open benchmark for evaluating AI models on observability and SRE tasks. And today, we're excited to share two updates: o11y-bench is now part of the Specialized Intelligence Index from Fireworks, the AI training and inference platform, and we're releasing o11y-bench 1.1, with updates that make grading more grounded, consistent, and reproducible.

Bringing o11y-bench to the Specialized Intelligence Index

The Specialized Intelligence Index brings together practitioner-built benchmarks for comparing how open, closed, and specialized AI models perform on domain-specific work, from healthcare and finance to cybersecurity and software engineering. It helps teams identify the best models for their workloads, understand where specialization adds value, and evaluate tradeoffs between quality, cost, and speed.

With o11y-bench now part of the index, SRE, platform engineering, and observability teams can explore how AI models perform on real observability workflows—from querying metrics, logs, and traces to investigating incidents and managing Grafana dashboards—alongside benchmarks for other specialized domains and use cases.

Run the benchmark with Fireworks inference

o11y-bench also now supports Fireworks as an inference provider in its default LiteLLM-backed agent. This means teams can evaluate both serverless models and models deployed within their own Fireworks accounts, using the same reproducible workflow available for other providers.

The integration handles Fireworks credentials, deployment names, and reasoning settings so benchmark runs are configured consistently. This makes it easier to reproduce results from the index or evaluate Fireworks-hosted models against your own requirements.

How o11y-bench evaluates models on real observability tasks

Observability deserves a place in model evaluation. During an incident, the hard part is rarely generating valid PromQL, LogQL, or TraceQL code. An engineer, or an agent, has to identify the relevant signal, use the right time range, connect evidence across telemetry types, distinguish a symptom from a cause, and sometimes modify persistent state without breaking something another engineer depends on.

o11y-bench evaluates those capabilities and more in a running Grafana stack. Agents use the Grafana MCP server to work with synthetic but realistic telemetry and complete 64 tasks across six areas:

Prometheus and PromQL
Loki and LogQL
Tempo and TraceQL
Multi-step incident investigations
Grafana API workflows
Dashboard creation, editing, and repair

The tasks are focused enough to grade, but still reflect real challenges engineers face in day-to-day observability workflows. For example, a query can be syntactically correct while selecting the wrong series, and a dashboard can still render even though its variables, data source, or saved query are wrong.

For this reason, o11y-bench starts with outcomes. It compares an agent’s claims with facts read from the same environment, and inspects the state the agent leaves behind. In other words, a convincing answer isn't enough; the results need to back it up.

Why consistency matters

Each o11y-bench task is run three times. The benchmark measures not just whether a model can solve a task once, but how consistently it performs that task across repeated attempts.

That consistency is especially important for observability. A model that reaches the right answer one time might follow a false lead or report an unsupported value the next. During an incident, that inconsistency can cost time and make it harder for teams to trust the results.

o11y-bench prioritizes Pass^3, the average benchmark score across three runs, while also reporting Pass@3, which captures whether the model succeeded at least once. Looking at both helps separate occasional wins from repeatable performance.

Strengthening the benchmark: what’s new in o11y-bench 1.1

Alongside o11y-bench's inclusion in the Specialized Intelligence Index, we’re releasing o11y-bench 1.1, which makes grading more outcome-based, reproducible, and resistant to answers that sound right without being grounded in the environment.

The improvements fall into three areas:

Grounded outcomes: Rather than rewarding a particular approach or phrasing, o11y-bench checks the outcome: whether the model’s claims match the facts, its saved changes are correct, and its behavior meets the task’s requirements. Reported numbers are checked against reference values, and cited traces are read back from Tempo to verify that they support the agent’s claims.
Deterministic grading: Dashboard edits are evaluated using before-and-after diffs of saved state. If a ground-truth query is empty or fails, the result is treated as a grading error rather than being allowed to influence pass rates.
Reproducible, calibrated evaluation: Saved facts let results be regraded without rerunning the agent, and scenarios remain consistent over time. A no-tools control failed all 64 tasks, showing that plausible answers alone cannot pass.

Because o11y-bench 1.1 uses updated grading and task calibration, its results are not directly comparable with 1.0. Version 1.1 establishes a new baseline for future comparisons.

Open benchmarks make better benchmarks

o11y-bench’s tasks, environments, grading logic, and results are all open source. That means anyone can examine how the benchmark works, reproduce its results, and help identify areas for improvement.

Adding o11y-bench to the Specialized Intelligence Index extends that work to a broader model-evaluation effort. We’re excited to see observability represented alongside other specialized domains, and to give engineering teams more evidence for evaluating the models they use in their workflows.

We also expect the benchmark to keep evolving. New models, agent harnesses, observability tools, and failure modes will expose gaps and create opportunities to make the evaluation stronger. As always, we welcome feedback and contributions from the community as part of that process.

To get started:

Compare models: Explore the o11y-bench results in Fireworks’ Specialized Intelligence Index.
Understand the benchmark: Review the tasks, categories, and detailed results at o11ybench.ai.
Run it or contribute: Visit the o11y-bench repository on GitHub.

Agent Observability FAQ: Understand the value of monitoring, evaluating, and improving your AI agents

Aishlee Kailey — Tue, 27 Oct 2026 14:32:18 +0000

Agents have quickly become part of the tech landscape as organizations embrace AI across their businesses. But embracing agents also means embracing the monitoring best practices that have long been part of engineers' workflows: tracking performance, catching errors, and responding to incidents

But it doesn't stop there. Agents also introduce new challenges. They can give different answers to the same question, be confidently wrong without throwing an error, and turn one request into a long chain of model and tool calls that’s hard to debug. A small prompt change can shift behavior as much as a code deploy. And because agents act on what they read, they open the door to new risks like prompt injection and leaked sensitive data.

It's a lot to learn, but that's why we built Agent Observability, now generally available, as part of our growing set of AI-enabled features in Grafana Cloud. It improves your agents by helping your teams:

Monitor performance and cost: Track latency, errors, token usage, and costs, and drill into individual conversations when something goes wrong.
Evaluate agent quality: Use built-in or custom evaluators to measure quality and detect issues like PII or prompt injection.
Test changes before you ship: Compare models, prompts, and other changes with offline experiments to catch regressions.
Build evaluations into CI/CD: Use experiments as deployment gates and track agent versions as you iterate.

Agent Observability was born out of our own challenges managing agents at scale, and it's the same tool we use internally today. In this blog, I'll address some of the common questions we get from customers.

Everything you want to know about Agent Observability

Monitoring, evaluating, and improving agents

Q: Why use Agent Observability to monitor my agents?

Agents are production workloads, and they need the same visibility as the services they touch. Agent Observability runs in Grafana Cloud alongside your metrics, logs, traces, and profiles, so you can tell whether a bad answer came from the model, the prompt, or a service the agent called. Monitoring, evaluations, guards, and experiments all live in one place, so you don’t need a separate tool.

Q: Can I monitor coding agents?

Yes. Integrations for Claude Code, Codex, Cursor, GitHub Copilot, OpenCode, Pi, and more send sessions to a dedicated Coding agents view, where you can track cost, tokens, and tool usage by user, repository, or branch. Want to try it outside of Grafana Cloud first? The open source local app keeps sessions on your machine, no Grafana account needed. See how to set up and monitor coding agents for more details.

Q: How do I know if a prompt or model change made my agent better or worse?

Agent Observability tracks agent versions automatically from changes to your system prompt, so you can compare quality, cost, and latency across versions. To catch regressions before you ship, run experiments against a baseline and use them as a CI gate.

Q: Do evaluations require an LLM judge?

No. You can use deterministic checks like regex or JSON schema validation alongside LLM judges, and mix both in the same evaluation.

Q: Can I see how multiple agents work together?

Yes. Each conversation shows a map of how your agents hand work to each other, even across services. If one agent’s output fails an evaluation, you can see which steps downstream were affected. See our docs on instrumenting agent dependencies and workflows to set it up.

Working with your existing stack

Q: Can I use my existing OpenTelemetry instrumentation?

Yes. Agent Observability accepts spans that follow the OpenTelemetry generative AI semantic conventions, so you don’t need the Agent Observability SDK. Keep exporting through Grafana Alloy or your OpenTelemetry Collector, and opt in your spans in with a single resource attribute. See our docs on instrumenting agents with OpenTelemetry to get started.

Q: Can I build my own dashboards or get alerted automatically if something goes wrong?

Yes. Agent Observability metrics land in Prometheus, so beyond the built-in views. You can also build custom dashboards and run ad hoc queries, plus set up your own alerts for cost spikes, latency SLO breaches, and more.

Cost and pricing

Q: Can I see what’s driving my agent costs, and where to cut them?

Yes. Cost and token usage are tracked per agent, model, and conversation, and you can tag spend by team, project, or environment. Coding agents can also break it down by user, repository, or branch. From there, you can spot calls that are using a pricier model than they need, track cache hit rates, and compare cost across agent versions. See our guide to optimizing cost and performance for more information.

Q: How is Agent Observability priced?

Agent Observability bills on three separate meters: generations (each model call your agent makes), tokens used by LLM-based evaluations and guards, and the standard metrics and traces your SDKs send. Free and Pro plans include 30,000 generations and 25 million evaluation and guard tokens each month. That token allowance is shared with other Grafana Cloud AI features. See Agent Observability pricing for current rates.

Security, privacy, and access

Q: Can Agent Observability block bad behavior, not just flag it?

Yes. You can create guards that sit on the request path before each LLM call, using regex patterns and PII detection to catch or redact sensitive data before the model ever sees it. You can also block risky tool calls.

Q: What data gets captured, and can I keep sensitive content out of it?

Agent Observability captures what your agent sends, including prompts, responses, tool calls, and token usage. The SDK can redact secrets before any data is exported, and guards catch PII before the model sees it. Coding agents send only metadata unless you opt in to full content.

Q: Can I control who sees conversation content?

Yes. Built-in roles let you separate access. For example, the Data Reader role shows analytics, traces, and evaluation results without access to conversations. Each Grafana Cloud stack is fully isolated, so use separate stacks when you need hard boundaries, such as dev versus production.

Q: Is Agent Observability available outside Grafana Cloud?

Agent Observability for agents you build runs in Grafana Cloud. For coding agents, there’s also a free, open source local app that keeps sessions on your machine with no Grafana account needed. When you want to observe coding agent usage across your team, connect it to Grafana Cloud.

Get started with Agent Observability

Agent Observability is available today in Grafana Cloud, including on the free tier, which comes with 30,000 generations and 25 million evaluation and guard tokens each month. To get started, create a free account and follow our guide to setting up Agent Observability.

Want to build a new agent? The open source Grafana AI SDK for Go gives you one interface across model providers, and connects to Agent Observability from the first call. Have a question we didn’t cover? Ask us in the Grafana community forums.

Spend less time on operational busywork and resolve incidents faster with Grafana Workflows

Michael Wettick — Wed, 21 Oct 2026 15:06:02 +0000

You're investigating checkout errors when someone asks who's coordinating the incident. Support needs an update, and another responder is looking for the runbook. You stop what you're doing to get everyone pointed in the same direction.

They're all reasonable requests, but each one takes your attention away from understanding the problem. Yes, your team has a process for handling incidents, but carrying it out is another job for someone who's already busy.

Grafana Workflows, available in public preview, turns your team’s processes into automated steps across Grafana Cloud and your other tools. You can even include Grafana Assistant in those steps to investigate an issue, summarize findings, or draft an update. For example, a workflow can create an incident channel, page support according to your team’s rules, and ask Assistant to prepare a briefing from the available evidence.

The same approach works with your typical day-to-day operations. Once you define the workflow, Grafana Cloud can prepare the next engineer's handoff or gather the findings for your weekly service review. You decide what should happen consistently and where AI can help make sense of changing information.

Get the incident response moving

To illustrate how your team can benefit from Workflows, let's walk through a major incident response process.

For starters, your team needs a place to collaborate and someone to coordinate the response. A runbook helps, but someone still has to find the on-call commander, open the channel, and share the right links.

Workflows include blueprints, prebuilt templates you can adapt to your team's process. For this type of project, you'd start with the major incident kickoff blueprint, which can automatically respond to an active critical incident in Grafana Cloud IRM, including an incident that has just been escalated to critical.

It creates a linked Slack channel, looks up your incident commander schedule, and assigns an on-call commander. It also posts your response instructions, bookmarks the runbook and incident page, and records the kickoff on the timeline. If nobody is on call, it leaves a note so the team knows to assign a commander.

Now responders have the incident record and runbook at hand, and the commander can start coordinating the investigation. The setup follows the process you've configured.

From there, you can build on that foundation. Suppose your policy is to involve support in every critical incident. Simply add a step that pages your support team in IRM and another that shares the incident link in their Slack channel. Support can get involved while engineers investigate, without waiting for someone to remember another handoff.

Note: Workflows is currently available under "More apps" in your Grafana Cloud navigation menu.

Put AI findings where people need them

Getting everyone together is just the beginning. The next person to join will need to know what's happening, what you've checked, and where to look next.

Extend the kickoff workflow with an ask Assistant step. Give Workflows the incident link, the affected service, and a specific task:

Review this incident and the available error and latency signals around its start time. Summarize the observed symptoms, link to the evidence, and suggest the next checks. Separate observations from hypotheses and flag missing information.

Assistant investigates using the available observability context and returns a briefing. Subsequent workflow steps post the briefing to the incident channel and add it to the incident timeline. The team can follow the evidence and ask better questions, and those who join later have a starting point for catching up.

At the same time, support has a different need. They have to turn technical findings into clear, customer-facing updates. An incident status update could trigger a workflow that asks Assistant to draft that explanation from the confirmed facts. A post message to Slack channel step sends the draft to the internal support channel, where the team reviews and edits it before sharing it externally.

This is where combining rules and AI becomes useful. You define the trigger, the task, and the destination. Assistant interprets the supplied context and returns a result that the next step can use. An engineer doesn't have to copy the answer out of a chat and put it in the right place.

You can also use AI to answer a specific question that determines the next step in your response. For example, when an alert fires, a workflow could ask Assistant to assess whether customers are affected and estimate how many. If the result exceeds a threshold you define, the workflow can declare an incident and page the SRE on-call. This focused assessment can run alongside a broader investigation into the root cause.

Give the next on-call engineer a head start

Soon after, the rotation changes, and the incoming engineer asks, "Anything I should know?"

Answering that question sufficiently can take some work. You remember the incident that's still open and fresh on your mind, but the incoming engineer also needs to know which alerts deserve another look. Gathering that context becomes one more task to finish before you can hand over the shift.

The on-call handoff report blueprint makes preparing that context part of the rotation change. It watches your selected IRM schedule and asks Assistant to review the team's recent incidents and alert groups. By default, it looks back over the past week, then posts a briefing to Slack with links to the relevant records.

The prompt asks Assistant to lead with what needs attention now: active incidents, recurring alerts, and follow-up work. You can adjust the time window and the report's focus to match your rotation.

The outgoing engineer can add anything that hasn't made it into the records. The incoming engineer gets a briefing to check and build on as they take over. That makes the handoff easier for both people, especially after a busy shift.

Come to the service review with the findings ready

The incident is now resolved, but the work doesn't stop there. Routine service reviews have their own preparation burden. For example, you may want to discuss whether latency is getting worse or error budgets need attention. First, someone has to open the dashboards, compare the numbers, and assemble an update for everyone else.

The weekly service health summary blueprint can have that first assessment waiting in Slack. Set it to run before the meeting for the services your team owns.

Assistant reviews the previous week's SLO status, error-budget consumption, and error and latency trends, along with relevant incidents and alerts. It produces a health assessment for each service and calls out risks for the coming week. The next workflow step posts the summary to Slack.

Your team comes to the review with an assessment to discuss, so you can spend the time deciding what needs attention.

Start with one process your team already repeats

Ready to get started with Workflows? Pick the part of your week that keeps requiring someone to gather context or coordinate the next step. Choose a starting point from the blueprints library, or describe your process to Grafana Assistant and ask it to build a workflow for you.

The workflow opens disabled, so you can review and test it before turning it on. The visual editor shows how the steps fit together. Run history lets you inspect their inputs, outputs, and errors, helping you check both the actions taken and the information passed between them.

For example, click on the button below to create your first workflow using an Assistant prompt to build a daily summary of incidents and firing alerts.

Grafana Workflows is available now in public preview. Use the getting started guide to try your first blueprint and learn more about this new feature in Grafana Cloud. Start with the incident setup you keep repeating, the handoff that always takes too long, or the service review you wish you had time to prepare. Make that one process easier, then build from there.

ObservabilityCON 2026: A guide to all the announcements from Grafana Labs

Grafana Labs Team — Tue, 20 Oct 2026 16:30:02 +0000

Today at ObservabilityCON 2026 in San Francisco, we unveiled a range of new Grafana Cloud capabilities designed to reshape cloud economics, bring richer context and deeper visibility across your systems and end-user experiences, and help teams move toward autonomous operations.

Check out the keynote to learn more about the latest updates to the open agentic observability cloud, and read on for a quick recap of all the news from ObservabilityCON 2026.

Autonomous operations with the help of AI

With Grafana Assistant in Grafana Cloud, you can bring more of your operational tasks together in one place so you can resolve incidents faster and get back to building. And you're doing it with AI features that automate much of the undifferentiated heavy lifting while still keeping you in control.

Assistant Investigations, generally available since late July, lets you run longer, deeper investigations that can start automatically from an alert or incident, and teams can run them from the platform, Slack, the Grafana Cloud MCP server, or gcx.

And today we're adding new AI features to help you take your automation even further:

Assistant Watchers (now generally available) are always-on agents that detect changes not captured by predefined alert rules. You describe what to monitor in plain language and select the relevant data sources. Your watcher then establishes a baseline from historical telemetry patterns and surfaces findings when those patterns change.
Grafana Managed Sandboxes (public preview) gives you a controlled cloud environment to ask questions about source code, make changes, and prepare pull requests from conversations in Grafana Cloud. It uses context from your repository code and live telemetry to propose potential fixes within a sandbox environment that a person on your team can then review and approve.
Grafana Pages (public preview) lets you build visualizations beyond any fixed constraints, making them closer to an app than a dashboard. You describe what you need, and Assistant builds a page in the canvas—all without you having to write any code. Existing visualization components still work, and Pages supplement them as the next evolution of Grafana.

With Grafana Cloud, you decide which parts of your operations you want to handle manually and which parts you want to hand off to Assistant.

To learn more about how you can get the most out of Grafana Cloud's AI capabilities, check out our blog posts on autonomous operations and Pages.

Cloud economics reimagined

In the AI era, where teams are managing more telemetry than ever, they need to be able to find the right log line in seconds. They also need to know they're only storing the telemetry that's worth paying for. That's why we're making major query performance enhancements to Grafana Cloud Logs, built on technology from our Logline acquisition, and giving you more precision with Adaptive Telemetry, so your costs scale with the value you get from your data.

Keep only the data worth paying for: Adaptive Telemetry updates

Grafana Labs’ Adaptive Telemetry suite optimizes cost across metrics, logs, traces, and profiles by recommending what to aggregate or sample, so teams keep the telemetry that matters. Customers have cut telemetry spend by 30% to 50% on average.

“Moving to Grafana Cloud and using Adaptive Telemetry cut our metrics costs enough to fund logs and traces as well," says Carl Eastman, Principal Software Engineer, Rapid7. "Our visibility went up while our costs went down, and dashboards and queries are faster and less noisy.”

New updates cover logs, metrics, and traces:

Adaptive Metrics auto-apply (generally available) continuously applies recommendations, while existing dashboards and alerts keep working unchanged. With cost-first mode, a built-in safeguard only applies recommendations that won’t increase your total active series, so optimization never turns into a cost spike.
Adaptive Logs Archive (public preview) routes logs to an object storage bucket you own. Get the cost savings from not putting those logs in Grafana Cloud, but make sure they still exist in low-cost storage for investigations, audits, or compliance requests. And with Adaptive Logs replay (sign up for the private preview), you can bring those logs back on demand.
Query traces dropped by Adaptive Traces (generally available) to sample aggressively without worrying about losing the one trace you need. Retrieve any trace by ID from the last 24 hours, even if Adaptive Traces dropped it, so cost savings never come at the expense of an investigation.
Volumetric sampling (generally available) gives you a more valuable sample than probabilistic sampling. It spreads your sampling budget fairly across services and request types, so low-volume services aren't crowded out. It also adjusts as your traffic changes. In early testing, it has delivered about 25% higher information density than probabilistic sampling at the same volume.
Diversity sampling (generally available) stores a unique combination of attributes such as service, route, and status codes every 15 minutes. When your environment changes, you don't have to write and update custom sampling logic to catch a new service or code path.

To learn more, check out our Adaptive Telemetry blog post.

Find the log line that matters faster: Full-text search in Grafana Cloud Logs

Built on technology from Logline, which Grafana Labs acquired in April 2026, Cloud Logs customers are gaining access to faster needle-in-a-haystack queries, at no additional cost.

This full-text search capability, now generally available, helps teams find specific log lines, such as those containing request IDs or customer IDs, without scanning massive volumes of data. Faster queries can also benefit AI-driven investigations, where agents execute a large number of queries to help find root causes.

“When I ran our first query on newly submitted logs after enabling Logline, the improvement was clear,” said Timo Kirzdörfer, OpenTelemetry & Grafana Cloud Admin, DATEV eG. “We went from processing nearly a terabyte of logs in three minutes to around 100 gigabytes in under three seconds. That represents an 80-90% reduction in the volume of logs searched for newly submitted logs, and we saw the results immediately.”

To learn more about full-text search, please visit our docs.

Context is king

As your systems grow more complex, you and your AI agents need a clearer picture of system health, user experiences, and what needs attention. Our latest Grafana Cloud updates bring more of that context together, expand visibility into the end-user experience, and help streamline everyday workflows.

Full-stack observability: more connections, richer context with the Knowledge Graph

Most teams have telemetry from across their stack, but connecting those signals to understand what's happening and how it affects users isn't always easy. That’s why we're rolling out new capabilities to advance full-stack observability across Grafana Cloud.

With expanded Knowledge Graph coverage and richer operational context, people and AI agents can better understand how systems fit together to investigate issues and find root causes faster.

A more connected view of your stack

The Knowledge Graph is the runtime context layer in Grafana Cloud that brings together your system's components, their relationships, and health insights in one connected view.

We're evolving the Knowledge Graph to be a platform capability, available in every stack, where you can curate context across Grafana Cloud. We’re also extending the full-stack picture, bringing more of your environment into the Knowledge Graph:

Frontend apps (generally available): Frontend Observability maps frontend apps and user-facing signals to the services and infrastructure behind them.
Synthetic checks (public preview): Alerts from Synthetic Monitoring checks now appear alongside service health, so you can spot issues and investigate faster.
Additional Google Cloud, Azure, and AWS resources (generally available): Cloud Provider Observability now maps more of your Google Cloud, AWS, and Azure resources for a more complete picture of your cloud infrastructure.
AI agents (public preview): Agent Observability maps your AI agents alongside their dependencies, so you can see how they connect to the rest of your system.
Microsoft SQL Server databases (private preview): Database Observability now discovers SQL Server databases and maps them automatically.

Together, these updates help teams investigate issues across more of their stack, from the user experience for their frontend apps to the services, databases, and infrastructure supporting it.

More operational context for people and AI agents

Knowing how your system fits together is important, but to investigate issues effectively, you also need context that’s unique to your organization, including which teams are responsible for which services and what you've learned from past incidents.

We're bringing more of that context into the Knowledge Graph, and making it available to both engineering teams and the AI agents they work with every day:

Service ownership (private preview): An integration with Grafana Cloud Service Center connects services to the teams who own them, on-call schedules, and runbooks, so following a dependency also shows who can act on it.
Custom entities and relationships (generally available): Define and map custom entities and relationships that are specific to your organization's products, business processes, and more. Use Knowledge Graph mode in Grafana Assistant to help configure these connections.
Operational context from past incidents (private preview): An integration with Grafana Cloud IRM connects actions and outcomes from previous investigations to relevant parts of your system, so you can build on what you’ve already learned.

The Knowledge Graph is also available as runtime context for AI agents (generally available). Whether you're using Grafana Assistant or bringing your own agent through gcx, this lets you and your agents work from the same shared context.

For a deeper dive on these new features, check out our full-stack observability blog post.

Digital Experience Monitoring: see where users struggle, not just when systems fail

Healthy systems don't always mean happy users. To understand how your applications are really performing, you need visibility into the end-user experience.

In Grafana Cloud, we've been building that foundation for Digital Experience Monitoring by bringing together Synthetic Monitoring and Frontend Observability, connecting proactive testing with real user signals. Together, they help teams understand not just whether an application is working as expected, but how it's actually performing for users.

At ObservabilityCON, we’re spotlighting new features to further build out Digital Experience Monitoring with Grafana Cloud, so you can further close the gap between system health and user experiences:

Session Replay (public preview): Watch a recording of what your users actually did in your application. Correlate when something breaks to the underlying telemetry so that you know exactly what happened.
Mobile Observability (private preview): Extend Frontend Observability to your iOS, Android, React Native, and Flutter apps using upstream native OpenTelemetry SDKs. See what mobile users experience, and connect it to the same telemetry as the rest of your stack.
Critical User Journeys (private preview): Define which journeys matter most, so every team works from the same definition of “critical” instead of improvising its own. Explore a journey end to end and go from a drop in conversion rate to the root cause in a single view.

You can learn more about Digital Experience Monitoring with Grafana Cloud here, and get started today by signing up for the Grafana Cloud free tier today.

Grafana Workflows: turn processes into automated steps

Coordinating an incident response can be so time-consuming that it pulls you away from actually understanding and fixing the problem. But with Grafana Workflows (public preview), you can focus on what really matters by turning your team’s processes into automated steps.

Grafana Workflows gives you event-driven automation that connects actions across Grafana Cloud and your other tools, with Grafana Assistant available for steps that need analysis or writing. Prebuilt blueprints cover major incident kickoff, on-call handoff reports, and weekly service health summaries.

For example, a workflow can create an incident channel, page support according to your team’s rules, and ask Assistant to prepare a briefing from the available evidence. And once you define the workflow, Grafana Cloud can prepare the next engineer's handoff or gather the findings for your weekly service review. You decide what should happen consistently and where AI can help make sense of changing information.

Expand what’s possible with Grafana: the Grafana Marketplace is generally available

The Grafana Marketplace is now generally available, giving teams more ways to extend Grafana with specialized, partner-built plugins.

Since launching as a pilot program earlier this year, the Marketplace has grown to include more than 30 plugins spanning security, networking, cloud cost optimization, and more. These paid plugins come with partner-backed commitments to ongoing maintenance, compatibility, and support, helping users confidently connect Grafana to the specialized tools they rely on everyday.

Grafana Cloud users can now try Marketplace plugins free for 14 days, with full functionality throughout the trial. Learn more in our blog post, and start exploring the Marketplace today.

Check out the ObservabilityCON 2026 homepage for all the latest updates from Grafana Labs' flagships observability event!

From production problem to proposed fix: Implement autonomous operations with Grafana AI

Maurice Rochau — Tue, 20 Oct 2026 16:30:02 +0000

You've just finished a deployment, and though your service is up, you start to notice some requests are getting slower. To address the lag, someone—or, more likely, some group of people on your team—needs to watch the dashboards, decide whether the change matters, investigate what happened, and turn the findings into a fix.

At each step you're gathering context—and often explaining that context again to the next person or tool. To make matters worse for the team responsible for the service, all this work is getting in the way of everything else they need to build.

With Grafana Assistant in Grafana Cloud, you can bring more of those tasks together in one place, helping you resolve incidents faster so you can get back to building. And you're doing it with AI features that automate much of the undifferentiated heavy lifting while still keeping you in control.

Assistant Watchers (generally available today as part of ObservabilityCON 2026) provides ongoing attention to the conditions you care about. Assistant Investigations (generally available) gathers evidence about what changed. And Grafana Managed Sandboxes (public preview as of today) uses that context to prepare changes in a controlled cloud environment, giving your team something concrete to review. In this blog post, I'll show you how you can use these features collectively for better observability outcomes.

Give the service ongoing attention

Consider a checkout service after a release. You want to know whether customers are encountering sustained errors or slower responses, while also accounting for the brief disruption you expect during deployment.

With Watchers, our always-on agents, you describe that concern and select the relevant data sources. Assistant then analyzes your historical telemetry to establish a baseline and calibrate the checks. From there, the watcher periodically assesses those signals and provides reporting along with the evidence behind its findings.

If a watcher finds a critical issue, it can automatically start an investigation with Investigations, our agent designed to help you go really deep into your incidents. That way, you can delegate recurring health checks to Assistant so your team can focus on other tasks instead. Existing alerts continue to cover known conditions that require dependable paging, while the watcher adds contextual assessment around the service.

The best part: With support for Google BigQuery alongside operational telemetry sources, your watchers aren't limited to traditional observability scenarios. They can also monitor business signals, helping your entire organization keep tabs on critical changes that could impact the bottom line.

Start investigating with the evidence attached

Of course, a notification is just the beginning of the response. The next questions are familiar: Which requests are affected? When did the behavior change? Is the database involved, or is another dependency slowing down?

Investigations can explore metrics, logs, traces, and profiles, test hypotheses, and assemble findings with supporting source queries. From there, your responders can inspect the evidence and steer the investigation as they learn more.

In our checkout example, suppose the evidence points toward repeated database calls in a request path. The useful handoff to a developer now includes the affected service, the symptoms, and the evidence supporting that hypothesis. The user has a starting point for checking the code.

Turn the findings into a proposed change

Managed Sandboxes, which give Assistant a controlled cloud environment to develop proposed fixes using repository code and live telemetry, carry that investigation context into your codebase. With the relevant repository connected and coding enabled, an agent can inspect the code, prepare changes, and run available checks in a Grafana sandbox.

For our example, the task could be to examine the database access pattern, prepare a targeted change, and add a regression test. Managed Sandboxes' repository-specific environments can provide the tools and dependencies needed to build and test the project.

The goal should be for Assistant to deliver a proposed pull request that the team can evaluate: what changed, why it addresses the suspected cause, which checks ran, and what still needs verification. So when you respond, the investigation is already underway, with supporting telemetry and, as the workflow progresses, a proposed fix to review.

Keep the decisions visible, and stay in control

A passing test is useful evidence, but your team still needs to confirm recovery in production as part of its release process. Your reviewers need to decide whether the diagnosis fits, whether the proposed change is appropriate, and how to deploy it. After rollout, they can use production telemetry to check the outcome and keep monitoring for recurrence.

That makes autonomy practical: delegate recurring observation, evidence gathering, and preparation of changes, while retaining clear decisions about what reaches production. You decide which parts you want to inspect, whether it's with Managed Sandboxes, Assistant Automations (to run recurring tasks), or Watchers. For some tasks, you might opt to hand it off to Assistant completely.

It also creates a useful way to judge the workflow. Did your responder spend less time gathering context? Did the developer receive enough evidence to act? Was the proposed change easier to review?

Start with one service, then scale

You can implement a similar level of automation with Assistant today. To start, choose a service your team already watches manually after releases. Define the behavior that matters, calibrate a watcher, and review its findings. Configure investigation escalations where appropriate.

Then, where Managed Sandboxes are enabled, connect a relevant repository and try a bounded task with a clear review outcome. Start with a problem your team understands. Use each run to improve the checks, context, and verification. Expand the work you delegate as the results earn your confidence.

After you collected some experiences and found the mode you want to work in, you can scale this to your liking. Assistant features are all interoperable; you can use Automations to set up Managed Sandboxes or Investigations, chat in Slack about the results of the latest incidents, and analyze all your alerts in the last 30 days periodically to improve the ones that don’t perform.

We’re excited to see what you can automate to help improve your operations.

Grafana Pages: Turn data into interactive experiences with the next evolution of Grafana

Maurice Rochau — Tue, 20 Oct 2026 16:30:02 +0000

Teams often spend as much time preparing information as discussing it: gathering charts, checking business reports, and finding the GitHub activity that explains a change. Every follow-up question can lead to yet another round of work.

That's why, as part of ObservabilityCON 2026, we're excited to introduce Grafana Pages, a new feature in Grafana Cloud that gives you a canvas to describe what you need and build a live app in minutes, based on your real data, context, and interactions. From new types of visualizations to slides and even to games, the only limitation is your imagination.

With Pages, currently in public preview, you can create interactive views and tools with Grafana Assistant, using the data you already access through Grafana. There's no code and no setup; no export, upload, or deploy steps. Unlike standard dashboards, which are fixed grids of panels, you simply describe what you need, refine it through conversation, and share it with your team to explore and return to as needed.

What Pages unlocks

Pages let you build trusted, reliable, and scalable (mini) apps in the environment you know and use with Grafana Assistant by your side.

Oftentimes, broad solutions might miss a use case that’s specific to your organization. But with Pages you can easily build and own your own information hub, whether it's an incident view, a runbook, a deploy tracker, an exec readout, or a customer-facing status page.

Let's look at some ways you can build Pages today.

Build on your existing data connections

Getting an internal tool working starts with connecting it to useful data. Pages builds on the data connections you already use in Grafana, giving you a starting point for bringing operational and business information together.

Put transaction volumes beside service performance, or build a release review around production trends and relevant connected context. Spend your effort shaping the experience around the question your team needs to answer, rather than getting bogged down trying to figure out all the visualization bells and whistles.

Let colleagues answer the next question

A shared view often leads to other questions you hadn't initially considered. One person might want to know about impacts on a specific region. A second might ask when the issue started, while a third wants to know which service is driving the change.

Pages can include filters, time-range controls, and drilldowns so colleagues can explore those questions themselves. Start with an overview, narrow it to a region or service, and inspect the supporting detail—all without asking someone to prepare another report.

Reuse the work; take the idea further

Build a page once, then return to it with a new time range for the next review. Refine it through Assistant as your team discovers which comparisons and details matter.

The remix feature in Pages gives a good idea room to grow. A colleague might build a useful deployment tracker; you might see how to extend it into a broader release review. Start from their Page, add your perspective, and turn it into something neither of you had to build from scratch.

Trust what you see

A page is only useful if people believe it. Every page keeps its sources, which are the queries behind each chart and number, so you and your colleagues can check where a value came from instead of taking Assistant's word for it. To validate a page you've built, open a source in Explore mode in Grafana Cloud to run the same query over the same time range, compare it with what the Page shows, and keep digging from there.

Pages also build on the access you already have. They work with any existing data connections you have configured in Grafana Cloud, and Grafana's permissions and data source policies still apply to every request. If a page can do something beyond reading data, like sending a Slack message or making an HTTP request, Grafana shows you the destination and values to review before it runs.

See what people are building

Early internal examples show the range of experiences people are creating with Pages, from business monitoring to release tools and visual storytelling.

In this section, we'll link to some of these projects so you can check them out. We've also embedded several of the pages directly into this post so you can interact with them here, too.

Run your operations

Pages can replace the tab-hopping that comes with triage, audits, and capacity reviews. Each of these examples gathers the data, ranks what matters, and lets you act on it.

Triage. Check the health of services across the (fictional) Astronomy Shop and Bank production namespaces. Triage surfaces SLO burn, error and latency regressions, and firing alerts. Pick the most urgent service and open an investigation from the same view.

Service health. See which services are burning or have spent their SLO error budget. Select a service to drill into its traffic, error rate, p95 latency, and firing alerts over the time range you choose.

Rightsizing. Compare Kubernetes CPU and memory requests against actual usage. Find workloads that are overprovisioned or starved, see the monthly cost of fixing them, and build a patch bundle of recommended request changes.

Alert Noise Audit. Find which alert rules produce the noisiest firing episodes. Classify them as flapping, always firing, broken, or night pager. Review fixes such as pending periods or keep-firing-for settings before you apply them.

Explore your data

A page can turn a data source into something you want to look at and click through, not only read.

Observatory. Request rates, error ratios, latency, service call paths, and recent orders for the Astronomy Shop appear as an interactive star map. The busiest services glow brightest, failing calls stand out, and you can follow a single order across the sky.

The Restless Planet. Recent USGS earthquake activity becomes a scrolling story for any time window you choose. It shows how often quakes occurred, where they clustered, and how their energy was distributed. Change the window and every sentence recalculates.

Tell the story

Pages aren't limited to dashboards. Use them to present results, because the numbers stay live while the format changes.

Week in Review. A weekly review of Astronomy Shop and Bank service health, SLOs, alerts, latency, cost, and customers. It's a navigable slide deck with presenter notes, ready for a Monday meeting.

Stack Wrapped. A story-style recap of your stack's week, covering requests, latency, alerts, SLO performance, regional revenue, and cluster cost. It's built to be shared, with tap-through cards.

Just for fun

Tail Run. Fly a game-style run through one minute-by-minute stretch of a production service's latency and error data. p99 latency sets the ceiling, and error rate and median latency set the floor. You can see how the service behaved while you play.

Start with one question your team keeps asking

You can start building Pages today. We suggest you choose a recurring task that sends someone across dashboards, reports, or connected tools. Identify who needs the result and what they should be able to explore. Then describe that to Assistant, using the sources you have connected.

For example, try this prompt in Assistant:

Create a store operations page using our transaction and payment-error data. Let me compare regions, select a store, and change the time range. Show missing data clearly and include links to the sources.

Try the first version, check representative values against the source, and refine the controls with the people who will use them. A useful next request might be: “Add a comparison with the previous week and put payment errors beside the transaction trend.”

You can even turn your most viewed dashboard into a page and see what it looks like with filters and drilldowns added.

Either way, start with something that saves your team one repeated round of preparation. Create a page and keep improving it as the questions change. Your first draft takes just a few minutes, from there on you can plan longer or shorter iterations and refinements.

And if you aren't already using Grafana Cloud, you can sign up for our forever-free tier and test Pages there, too!

Break the link between data volume and your observability bill with Adaptive Telemetry

JK Basra — Tue, 20 Oct 2026 16:30:02 +0000

In the AI era, where telemetry is generated by every model call, tool invocation, and agent step, organizations need more help than ever when it comes to navigating the economics of observability.

For a lot of teams, the default response is a "data diet" of shorter retention, which leaves services or environments out of collection, or basic sampling, which is more guesswork than strategy. Either way, you end up cutting to control the bill, and you risk losing the very data you need during an incident.

Your costs shouldn't scale with how much data you ingest. Instead, they should scale with the value you get from it, so you only pay to keep the data that earns its place.

That’s what Adaptive Telemetry is built for, and organizations such as Rapid7 are already seeing the results.

You can already use Adaptive Telemetry for all your observability signals, and we're continuing to build on that foundation so you can get even more precise about which traces you keep, improve your logs economics, and get more from Adaptive Metrics with less toil. Here's how that looks in practice.

Keep only the traces that matter, and cut volume by 90%

Traces are where the data diet hurts most. Many teams sample with a fixed percentage, which treats a slow checkout request the same as a routine health check. Cut the percentage and your bill drops, but so does the chance that the trace you need is still there when something breaks.

Adaptive Traces already handles the hard part (tail sampling) without you running a pipeline. Three newer capabilities make retention even more precise.

Automatically keep your most valuable traces, without the policy toil

Volumetric sampling gives you a more valuable sample than probabilistic sampling does. It spreads your sampling budget fairly across services and request types, so low-volume services aren't crowded out, and it keeps adjusting as your traffic changes. You set one number (your sampling percentage) without writing complicated policies or tuning them by hand. In Grafana Labs' testing, it delivered about 25% higher information density than probabilistic sampling at the same volume.

Keep at least one trace from each footprint, so there’s always something to reference

Diversity sampling makes sure you always have something to look at. Every 15 minutes, it stores a unique combination of attributes such as service, route, and status codes. When your environment changes, you don't have to write and update custom sampling logic to catch a new service or code path.

Sample aggressively and still retrieve the exact trace you need

The ability to query traces dropped by Adaptive Traces is your safety net. Adaptive Traces keeps every dropped trace queryable by trace ID for 24 hours after ingest. If you have the trace ID from a log line, a support ticket, or a testing framework, you can retrieve that specific trace even if Adaptive Traces dropped it. Dropped traces don't appear in search or aggregations, so it's a lookup, not a replacement for sampling the traces you want to search. That's what lets you sample aggressively with confidence.

Together, you can use these features to cut trace volume significantly without wondering whether you cut something you needed.

Volumetric and diversity sampling, along with the ability to query dropped traces, are all generally available.

Improve your logs economics without giving up the data you might need

Logs add up fast, and a lot of what you store never gets looked at. Adaptive Logs already finds the log patterns nobody queries and trims them on the way in. Now, two new capabilities help you cut earlier, and with more confidence.

Drop unwanted logs at the source and cut egress and network overhead

Fleet-managed Adaptive Logs uses the same drop rules and exemptions you already use and applies them directly to the collectors you already run, through Fleet Management. Matching logs are filtered out before they ever leave your environment, so you save on egress and network overhead on top of your storage savings.

Adaptive Logs creates and keeps the collector pipelines in sync for you, so there's nothing extra to write or maintain. Pattern-based recommendations still run in Grafana Cloud for everything that remains.

Keep what you might need, retain it in low-cost storage

Teams are understandably cautious about dropping logs. After all, a dropped line can't be recovered, so it's understandable when teams hold back. Archive and replay changes that.

Archive routes logs to your own low-cost object storage instead of dropping them. Replay brings them back on demand when you need them for an investigation, an audit, or a compliance request. You keep the savings from not indexing low-value logs, without permanently losing data you might need later.

That's the shift: lower costs, with a way back to the data you might need.

Fleet-managed Adaptive Logs is in public preview. Archive is in public preview, and replay is coming soon. Sign up for the private preview to get early access.

Get more from Adaptive Metrics, with less toil

Adaptive Metrics is already good at finding the waste. It looks at how your metrics are actually used across dashboards, alerts, recording rules and queries, and it recommends aggregating the ones nobody needs at full detail into lower-cardinality versions. The catch is that someone has to review each recommendation, and for platform teams, that review quickly becomes its own chore. Rapid7's Eastman knows the feeling:

“Before, finding waste meant someone manually digging through metric cardinality and log volume, and the job never ended because new services kept adding more," Eastman said. "Adaptive Metrics and Adaptive Logs now do that analysis for us... Our engineers spend their time on the few rules that need human judgment... rather than the whole long tail.”

Apply savings automatically, without reviewing every recommendation

Auto-Apply takes the review off your plate. Turn it on for a segment, and new recommendations are applied automatically as they're generated, with no human in the loop.

You choose where it runs, and you can exempt the metrics you never want touched. Use coverage-first mode to keep pace with usage automatically, or cost-first mode if you’d rather guarantee your series count never increases

Auto-Apply is generally available. For the full walkthrough, see our recent blog post on the topic.

Ready for the next wave of telemetry

AI will bring more telemetry, but it doesn't have to bring more cost and more noise with it. Trim what nobody uses, keep the signals that matter, and know the data you might need isn't lost for good. You don't have to cut retention, leave services out of collection, or sample on a hunch.

That matters beyond the bill. When your costs follow the value of your data, the savings are yours to reinvest in broader coverage and better insight, instead of living on a data diet. Rapid7 is a good example. Here’s how they describe it:

“We’re reinvesting the savings in observability coverage rather than just banking them," Eastman said. "Every signal has to earn its place, and the savings pay for signals that do.”

Whether it's metrics, logs, traces, or profiles, Adaptive Telemetry helps you keep what's valuable as your data grows.

Adaptive Telemetry is available in Grafana Cloud. Try it today, and read our docs to learn more!

Full-stack observability in Grafana Cloud: More connections, richer context, and faster investigations for users and agents

Sarah Constant — Tue, 20 Oct 2026 16:30:02 +0000

Most teams have telemetry from across their stack, but connecting those signals to understand what's happening and how it affects users isn't always easy. Critical context about service dependencies, ownership, and past incidents often lives across tools, runbooks, and conversations between team members.

That’s why, today at ObservabilityCON 2026, we’re announcing new capabilities that advance full-stack observability across Grafana Cloud. Powering this experience is the Knowledge Graph, the runtime context layer in Grafana Cloud that brings together your system's components, their relationships, and health insights in one connected view.

We're expanding the Knowledge Graph to cover more infrastructure, applications, and user-facing signals, along with a broader range of operational context, including your service catalog, the teams on call, and lessons captured during past incidents.

Together, these connections give people and AI agents the same shared context, built on open standards, about how a system fits together, helping them to troubleshoot and find root causes faster.

“We've found that the hardest part of an investigation is rarely a lack of data. The challenge is connecting service ownership, dependencies, operational knowledge, and customer impact across a complex ecosystem,” said Neil Wilson, Senior Director of Software Engineering at LexisNexis Risk Solutions. “Entity graph relationships [on the Knowledge Graph] help make those connections visible, giving engineers and AI systems the context they need to move from data to understanding to action."

Think of Grafana Cloud as your SRE specialist: whether you work with Grafana Assistant or bring your own agent through gcx, you want that specialized expertise available when you need it, without spending the first half of an investigation reconstructing how everything fits together. That's where we're taking full-stack observability: towards faster, more reliable, and more cost-effective answers, powered by the Knowledge Graph.

A connected view of your system: making the full stack even fuller

We're evolving the Knowledge Graph to be a platform capability, available in every stack, where you can curate context across Grafana Cloud.

Traditionally, the Knowledge Graph has supported Grafana Cloud Application Observability, Kubernetes Monitoring, Database Observability, and Cloud Provider Observability. These solutions automatically discover their respective entities, such as services, pods, and nodes, and map these entities and their relationships onto the Knowledge Graph. They then monitor entity health and add these insights onto the graph, highlighting everything from deviations from dynamic baselines to configuration changes.

Now we're extending that full-stack picture, bringing more of your environment into the Knowledge Graph:

Frontend apps (generally available): Frontend Observability maps frontend apps and user-facing signals to the services and infrastructure behind them.
Synthetic checks (public preview): Alerts from Synthetic Monitoring checks now appear alongside service health, so you can spot issues and investigate faster.
More Google Cloud, AWS, and Azure cloud resources (generally available): Cloud Provider Observability now maps more of your Google Cloud, AWS, and Azure resources for a more complete picture of your cloud infrastructure.
AI agents (public preview): Agent Observability maps your AI agents alongside their dependencies, so you can see how they connect to the rest of your system.
Microsoft SQL Server databases (private preview): Database Observability now discovers SQL Server databases and maps them automatically.

With these updates, you can investigate issues across more of your stack, from a user-facing app to the services, databases, and infrastructure behind it.

Go beyond telemetry with richer context in the Knowledge Graph

We're bringing more of that context into the Knowledge Graph, and making it available to both engineering teams and the AI agents they work with every day.

Service context: connect services to their owners with Grafana Cloud Service Center

While your telemetry describes what’s actually running, a service catalog describes your system in terms your organization understands. Connecting the two gives an investigation a route from a deployed component to the service it belongs to, the team responsible for it, and the guidance they maintain.

The new Knowledge Graph integration with Grafana Cloud Service Center, now in private preview, brings these connections onto the graph, so you can see service ownership, on-call schedules, and runbooks alongside the components they support.

For example, let’s take Everyday Checking, an imaginary banking app powered by account management and card payments services. If account creation slows down, you need to know which deployment is affected, which team owns it, who’s on call, and where to find the account creation runbook. Having all that information connected in the Knowledge Graph gives an engineer or an agent a much more useful place to start.

Team context: add custom entities with Grafana Assistant

Your team knows things that no automatic discovery process can infer. Account management and card payments might belong to the Everyday Checking app. A group of services might support a specific online product or business process. You should be able to make those relationships available to everyone in your org via the Knowledge Graph.

That’s why we built custom entities and relationships, now generally available. They let you define rules based on your telemetry to create the entities you need, connect them to the rest of the Knowledge Graph, and attach health insights. For example, if multiple services share the label product_line="everyday-checking", you can use that label to group them under a new Everyday Checking product entity.

You can also use Knowledge Graph mode in Grafana Assistant, now generally available, to help define these custom entities and relationships. Ask how to model your product, write a custom rule, or investigate a missing connection, and Assistant can guide you through the process.

Operational context: give the next investigation the benefit of the last

Some of your most useful operational knowledge is created when something goes wrong: someone tests a hypothesis and rules it out; a restart appears to help, but the errors return; a configuration change finally restores service. By the end of the investigation, the team has learned far more than the final root cause.

Through an integration with Grafana Cloud IRM, we’re bringing those lessons onto the Knowledge Graph as operational context, now in private preview. This goes beyond attaching an incident summary to a service. We’re extracting hypotheses, actions, and their outcomes from the evidence captured during investigations, and connecting them to the relevant parts of your system.

To revisit our Everyday Checking example, imagine the team tried a restart, saw the errors return, then restored account creation by reverting a database connection pool setting. The next investigation could use that history to check whether the same conditions apply, and whether another restart is worth trying. The earlier incident provides a lead to test against current telemetry.

Runtime context: give agents and people the same starting point

Each connection surfaced in the Knowledge Graph makes it possible for teams to ask more useful questions. Which services support this product? Which deployment is affected? Who owns it? What did we try the last time this happened?

People can explore those connections in Grafana Cloud. And now, with the Knowledge Graph available as runtime context for AI agents, a capability that’s generally available, Grafana Assistant and your own agents connected through gcx can work from that same shared understanding of your system.

“We had a latency alert fire, then a second alert in our downstream HPCC system. The real issue turned out to be a call to an external vendor that exhausted the connection pool at our database layer,” said LexisNexis Solutions’ Wilson. “Assistant walked down the Knowledge Graph and got us to that root cause a lot quicker, and it kept teams off paths they had already started down.”

Our new Cypher query tool also lets agents ask the Knowledge Graph direct questions about how your system connects. Cypher is a standard graph query language: you describe a pattern, such as “every service that calls this database and the pods it runs on,” and get back each match with its relationships and current health. This helps agents build on context already available to them rather than reconstruct from raw telemetry each time.

Get started today

With more of your system and operational knowledge connected through the Knowledge Graph, you and your AI agents have a clearer picture of how everything fits together. That means you can spend less time piecing together context and find root cause faster.

Ready to get started? Ask Grafana Assistant: “How can I get more from my Knowledge Graph?” and get specific recommendations based on your stack. Note: some changes may require admin access.

You can also fill out this form to access the private previews mentioned throughout this post, including mapping Microsoft SQL Server databases to the Knowledge Graph, and learn more in our Knowledge Graph documentation.

Expand what’s possible with Grafana: The Grafana Marketplace is now generally available

Sofia Neroda — Tue, 20 Oct 2026 16:30:02 +0000

Earlier this year, we launched the Grafana Marketplace, a platform that gives independent software vendors, systems integrators, and developers in the Grafana community a new way to sell and distribute plugins built for Grafana.

The Marketplace builds on our existing plugin catalog by enabling third-party partners and developers to earn revenue from their plugins and invest further in their development. For end users, these paid Marketplace plugins come with guarantees for ongoing updates and support, helping ensure the integrations you rely on remain secure and reliable over time.

Since launching the pilot program in April, the Marketplace has grown to include more than 30 custom, partner-backed plugins for high-demand data sources and apps, ranging from cloud cost optimization tools to security solutions from Rubrik and Cisco.

Today at ObservabilityCON 2026, we’re excited to share that the Grafana Marketplace is generally available. And to make it easier to explore, any Grafana Cloud user can now trial Marketplace plugins free for 14 days.

Why the Marketplace matters

The Grafana plugin catalog includes more than 300 plugins built and maintained by Grafana Labs, our partners, and the community. This ecosystem gives users more ways to connect Grafana to the tools and data sources they rely on, while giving developers the freedom to extend Grafana for new use cases.

We want developers to be inspired to keep building on Grafana, while our users have more options for monitoring their infrastructure and managing their Grafana instances, including apps and data sources for specialized use cases.

But building and maintaining a production-grade plugin takes a lot of work. Until recently, independent developers and systems integrators had no direct way to fund that work through the Grafana plugin ecosystem.

The Marketplace introduced that commercial model. By giving partners a way to earn revenue from their plugins, they can invest in ongoing development and support while bringing more specialized solutions to Grafana Cloud and Grafana Enterprise users across observability, security, IoT, and beyond. The result is more production-ready data sources and apps reaching the catalog, backed by the ongoing support users need to rely on them.

Importantly, the Marketplace adds to our existing plugin ecosystem; it doesn’t replace it. Community and open source plugins remain a core part of Grafana, and developers can continue to build and publish free plugins just as they do today. The Marketplace provides another path for developers who want to sell their plugins and take on the corresponding support and maintenance commitments.

Explore the growing Grafana Marketplace

Since we launched the pilot, the Marketplace has grown to cover more than 30 high-demand data sources and apps, with options spanning networking, security, storage, virtualization, collaboration, cost management, industrial infrastructure, and more.

Below are some examples of the Marketplace plugins available today:

Network and security: Cisco Meraki, Rubrik Security Cloud, Zscaler ZIA and ZPA, and Akamai data sources from Crest Data.
Storage and virtualization: NetApp ONTAP, Dell PowerScale, and VMware vSphere from Crest Data, and Proxmox VE from Ridgeline Technologies.
Productivity and collaboration: SharePoint, Citrix Cloud, Tableau, and Zoom from Crest Data, and end-to-end Microsoft Teams observability from Phenisys.
Cost and compute: Kubecost and Nutanix from Integration Plumbers.
Industrial infrastructure: Helin Connect Data Hub data source from Helin Data.
Manufacturing quality analytics: A full statistical process control (SPC) suite and 3D CAD visualization from KensoBI.

Marketplace plugins run inside your own Grafana instance, and Grafana Labs reviews every listing before it goes live. All plugins must be cryptographically signed and must encrypt sensitive data. Partners are contractually obligated to remediate critical vulnerabilities within seven days and high-severity vulnerabilities within 30 days.

While Grafana Labs handles billing and payment, support for these plugins comes directly from the partners who built them. We vet every partner and each one is contractually committed to keeping its plugins compatible with supported Grafana versions.

Try Marketplace plugins free in Grafana Cloud

Every Marketplace plugin comes with a 14-day free trial and full functionality from the first click. Find these plugins in the catalog, or in the Plugins tab of your Grafana Cloud stack, and start a trial.

If you're a contracted Grafana Cloud customer, please reach out to your account team to set up your trial.

We’d love to hear from you

The Grafana Marketplace is still early, and we want to keep building it the Grafana Labs way: with your input. If a technology you rely on isn't covered yet, reach out at plugins-marketplace@grafana.com.

And if you’re a developer interested in contributing your plugins to the Marketplace, please submit this form.