A teammate asks why you think a deployment caused checkout failures. You have an error-rate chart in your dashboard, a query that narrows the failures to one payment provider, and an explanation you've been exploring with Grafana Assistant. To bring your teammate up to speed, you start collecting screenshots and query links, then write out how they connect.

That explanation is part of the analysis. Your teammate needs to see what supports the conclusion and where you're still making an assumption. As findings come from both your own exploration in Grafana and from AI-assisted work, having one place to examine them together becomes key.

Grafana Notebooks help you do just that, bringing charts, queries, and findings from Assistant together in one working document, so teammates can follow your reasoning and assess next steps. In this post, we’ll use the checkout example above to show how this works in practice, then look at how Notebooks can support rollout decisions and post-incident reviews.

How Notebooks work: develop your analysis in a working document

Through conversations with users, we've heard a recurring need for a lightweight place to work through a question and share the results. Notebooks give that work room to develop inside Grafana, from an initial observation to a recommendation someone can assess.

You can start with a panel you've already found. Select Add to notebook from its dashboard menu, or from the Add menu in Explore, and choose a new or existing notebook. Then add your explanation and reorder the blocks as the argument takes shape. Your edits save automatically.

Assistant can contribute to your working document, too. Add a visualization generated in an Assistant conversation directly to a notebook, whether you're using the side pane or Workspace. Once you're in the notebook, open the Assistant side pane to ask for changes to a visualization or help refining the document. You can also edit the content directly as your analysis develops.

Work through a question and share your conclusion

Let's revisit the checkout example. Your notebook starts with the error-rate panel and a question: could the latest payment-service deployment explain the increase?

Alongside the panel, you bring in an Explore query that narrows the errors to requests to one payment provider. You can refine that query in the notebook without changing the source dashboard, and can also set an absolute start and end time on the error-rate panel to keep readers focused on the period around the deployment.


You ask Assistant to chart payment-provider timeouts before and after the deployment, then add the visualization directly from the conversation to your notebook. Within the notebook, you open the Assistant side pane and ask, "Break this chart down by payment provider." In this example, the breakdown shows an increase for the provider you identified in Explore. You add a detail from the deployment notes: the release changed that provider's timeout setting. You explain why that connection deserves testing.

Before sharing, you put the suspected cause and proposed test near the top, with the supporting charts and explanation below. A reader can see what you're recommending, then follow the evidence. The next step is to reproduce the behavior with the previous and current timeout settings.

You share the notebook link with the payment-service team. A teammate with access can read the analysis and examine the queries as they plan the test. Their next action happens in the service's test environment, informed by the work you've brought together.

Support rollout decisions, post-incident reviews, and more

The same approach works for questions outside troubleshooting. For example, after a deployment intended to improve performance, you could bring together latency and resource-usage charts, explain the differences you observe, and document any trade-offs. Share the notebook with the service owners to help them decide whether to expand the rollout or make another change.

For a post-incident review, a notebook can bring relevant queries together with an account of what happened. Place your explanation beside the evidence so readers can understand the conclusions before discussing follow-up work. A finding about repeated timeouts, for example, might support a proposal to change retry behavior or add a load test.

Start with a question your team needs to answer

For your first notebook, choose a question you're already working through in Grafana. Add the evidence you've found, explain what you think it means, and make the next question or recommendation clear. Share it with the people whose expertise you need.

We're also exploring richer collaboration in Notebooks, including comments and live editing. Our ambition is to give people and agents more ways to develop their work together, with the evidence and discussion in the same place.

Notebooks are in public preview in both Grafana Cloud and self-managed versions of Grafana, as of the Grafana 13.3 release. Grafana Cloud access is rolling out over the coming weeks. 

For self-managed Grafana, follow the Notebooks guide to enable the dashboard.notebooks feature toggle. Once available, create a notebook from the Notebooks page. You can also start with findings from an Assistant conversation.

Read the Notebooks guide for more details on setup and examples of adding content from dashboards, Explore, and Assistant.