groundcover brings Observability for Production AI to Microsoft Azure at Microsoft Build

Microsoft Build runs June 2–3 in San Francisco, groundcover's West Coast home, and the team is at booth g-222 for the full event running demos and releasing Agent Mode for Azure. In-person or virtual attendees can schedule time here.

Agent Mode launched for groundcover customers widely in March. This release extends it across the full Azure ecosystem, giving every engineering team managing Azure infrastructure access to what is essentially a groundcover expert user inside of the platform. This expert user, Agent Mode, is purpose-built to investigate production incidents and analyze infrastructure in partnership with an engineer.

An SRE inside your cloud, not someone else’s

When something in your environment breaks, Agent Mode works with you to investigate. It pulls logs, correlates traces, and works through telemetry conversationally. Working alone, what takes an engineer an unbounded amount of time for dashboard navigation, Agent Mode helps deliver in an instant. Colloquially, groundcover customers say they are able to identify the cause of performance issues before they even enter the UI nearly half of the time. 

Because groundcover deploys with a bring-your-own-cloud (BYOC) model, every investigation runs inside the customer's own infrastructure. Logs, traces, and production telemetry stay there. From a privacy point of view, there’s little if any concern with what data moves through groundcover.

That’s beneficial because deploying the eBPF sensor also expands what Agent Mode can see compared to contemporary telemetry practices. eBPF surfaces performance data from applications that were never manually instrumented with OpenTelemetry. In most production environments, that covers the majority of running applications, so teams don't need complete instrumentation coverage before Agent Mode delivers value.

Why AI for observability 

Agentic development platforms and foundational model providers have gotten good at making deployment sound straightforward. How to build agents that systematically improve a team's reliability practices is the question that's mostly gone unanswered.

groundcover built Agent Mode with several premises in mind, but with one foundational one: an agent earns its place in observability by processing data volumes faster than any engineer could on their own, and surfacing insights that help platform teams act on their biggest problems before those problems compound. The interface is the conversational front end engineers already know from working with LLMs, scoped entirely to telemetry emitted by applications, infrastructure, and AI, and equipped with tools to act on what it finds rather than just describe it.

Querying production data without giving it away

Teams coming from traditional SaaS observability carry a familiar constraint: querying your own data means routing it through someone else's cloud first. SaaS customers absorb the egress cost, accept the vendor's data terms, and inherit whatever security exposure follows. Production data isn’t limited to your production instance. 

By comparison, Agent Mode queries telemetry that never leaves the customer's architecture. Engineering teams use natural language access to understand production data, privately, at their own compute cost, without a third-party routing requirement embedded in the workflow. Privacy and session controls are configurable at the workspace level. groundcover customers are already running this in production, and their teams use Agent Mode's analysis without security caveats because the data stays inside their own cloud.

Pricing

Every groundcover customer on Azure gets Agent Mode automatically. Pricing is based on node count, a measure of compute, not data volume or query frequency, with no token markup for Agent Mode. Teams that have already negotiated commercial terms with Microsoft for Claude usage pay those same rates when Agent Mode runs inspections and surfaces recommendations.

Working outside the platform UI

For teams that prefer not to run queries from within the groundcover application, the platform data and tools are available through the groundcover MCP server. MCP gives teams access to private production telemetry, and Agent Mode skills, in whatever deployment structure fits how they already work.

Next steps

Agent Mode is the coordinator, inspector, and domain expert that Azure platform teams have missed while AI costs crowded out the budget and time for proper observability. Now, the data stays in their cloud, and the analysis is ready when they need it. 

groundcover is committed to helping our customers support their entire ecosystem, and dedicated to serving Observability in a deployment and commercial model that works actually delivers value to them commensurate with their spend. That’s why we want to work with you to understand and to design your observability solution. 

Schedule time with us and our solutions engineers to design this together. Also make sure you register for the upcoming webinar with veteran tech journalist, Chris J. Preimesberger and VP (marketing) Chris Churilo on the evolving role of AI workloads in observability platforms.