groundcover, the world's leading bring-your-own-cloud, eBPF and OpenTelemetry-native observability platform, today announced a major expansion with groundcover Dashboards, adding natural-language dashboard and widget creation built on its new query language, gcQL, alongside a deeper set of visualization and governance capabilities. The release runs counter to widely held industry assumption that AI agents make dashboards obsolete by moving everything into chat.
Across observability and developer tooling, there's dominant narrative that conversational interfaces will replace dashboards, and that user interface becomes secondary once an agent can answer questions directly. groundcover is arguing more specific version: dashboards are dead for agents, and very much alive for humans.
An agent doesn't have eyes. It cannot glance at a screen during an incident or notice that a graph looks wrong. But most people running production are still human, and chat thread is poor substitute for persistent, shared view of a system. It isn't shared, it doesn't persist, and it isn't there in middle of night when someone else is on call. What agents removed wasn't need for dashboards. It was hours it used to take to build one.
"The consensus right now is that agents make dashboards obsolete, that everything moves into a chat window," said Shahar Azulay, CEO and co-founder of groundcover. "For an agent, that's actually true. But dashboards aren't dead for people running production; they just need to be better, faster to build, easier to reason about, and ready in seconds instead of hours. That's what we spent this quarter on, and our customers are building more dashboards than before, not fewer."
What's new includes dashboards built by asking, not clicking. Ask Agent Mode for dashboard and get back complete, populated board on first pass, built faster than creating one manually. Describe outcome you want, such as health of workload, and agent determines which panels are needed, pulling in infrastructure usage, error traces, and error logs as needed. You can sketch dashboard on whiteboard or napkin, snap photo, and drop it into Agent Mode to have it built out automatically. There's no need to describe it in words at all.
Dashboards that are cheaper to build. Natural-language dashboard creation is now default flow, with manual start-from-scratch creation still available. Users can also generate individual widgets by describing them, preview them in existing dashboard without rebuilding it, or add visualization directly from Data Explorer. New dashboard catalog offers pre-built templates for CI/CD, databases, RabbitMQ, AWS cost, APM, and more, all previewable against customer's own data.
Keeping telemetry inside customer's own cloud rather than vendor's is becoming baseline requirement for any data-heavy operation. Dashboards make that requirement visible, but visualization is only as good as pipeline feeding it, and groundcover keeps that pipeline inside customer's own infrastructure by default.
Dashboards that are more capable. groundcover added configurable thresholds on time series charts, displayed as lines or filled ranges with customizable labels, particularly useful for teams constructing SLO dashboards. New tree map visualization shows where volume, errors, latency, or cost are concentrated. Stats widgets now support conditional color formatting and configurable decimal precision, while time series charts have gained new Y-axis controls, including scale, minimum and maximum bounds, always-include-zero, and log scale. Widget-level sharing lets user copy link that opens dashboard scrolled to and focused on specific widget, useful on dashboards containing 40 or 50 widgets. Hovering over table row surfaces drill-down into Explore, either for aggregate analysis or underlying rows.
Dashboards that are governable. groundcover reworked dashboard variables with new interface and advanced settings, including cross-source key mapping, so single variable can resolve across data sources with mismatched labels, such as cluster_id in logs and cluster.id in metrics. Dashboards can now be tagged for organization, with team ownership and permissions in progress to help ensure dashboard someone built can't be deleted by someone who didn't build it. Version history with revert and synced crosshairs across widgets are both included.
Teams migrating from Datadog will find that threshold visualizations and tree maps, two components with no prior equivalent in groundcover, now have parity, addressing common reason migrated dashboards previously broke.
"These pull in the same direction," said Orr Benjamin, VP of Product Management at groundcover. "Every visualization we add is one the agent can reach for. When someone asks for a view of error rate against deploys, the quality of what comes back depends entirely on what the dashboard layer can express. Building better primitives is how you get better agent output. The two aren't a trade-off, and treating them as one is the mistake."
groundcover Dashboards, gcQL, and capabilities described above are available now to all groundcover customers.
About groundcover
groundcover is the full-stack observability platform for engineers & agents, allowing teams to monitor everything they run in their cloud, without compromising on data, control, or cost. It's deployed inside the customer's cloud premises and provides complete visibility into applications, infrastructure, and AI workloads without operational overhead. The platform offers unlimited data coverage at a fraction of the cost of legacy observability tools. Founded in 2021, the company is backed by leading global investors, including One Peak, Morgan Stanley Expansion Capital, Zeev Ventures, Angular Ventures, Heavybit and Jibe.
