How is your enterprise tracking AI agent telemetry? Groundcover thinks it should never leave your cloud

groundcover announced a $100 million funding round led by One Peak, lifting its total capital to $160 million. The observability startup, which brands its name with a lowercase g, says it has more than 250 paying customers. The raise lands as enterprises start treating AI agent telemetry as a first-class monitoring problem rather than an extension of traditional APM.

The product thesis is blunt: agent telemetry should stay inside the customer’s cloud. That framing targets enterprises that cannot send prompts, tool calls, traces, or decision logs to a third-party SaaS control plane without breaking data-residency, security, or compliance rules. Observability for agents is less about dashboards alone and more about where spans, metrics, and event streams are collected, stored, and queried.

For engineers building multi-agent systems, that constraint matters immediately. Agent stacks produce high-volume, high-cardinality signals—tool invocations, model routing choices, retries, cost attributes, and failure modes that look nothing like classic HTTP request traces. If the observability path requires exporting that stream off-prem, builders hit legal review, network egress cost, and delayed incident response. An in-cloud collection model shrinks that friction and keeps debug data next to the workloads that generated it.

The market context is a fast-growing AI agent observability category where buyers are still sorting which products they actually need. Traditional observability vendors and newer agent-focused tools are competing for the same budgets, but enterprises remain unsure how much of the stack is logging, how much is tracing, and how much is policy or safety monitoring. groundcover’s $100 million raise and 250-plus customer base put a clear capital and commercial signal behind the in-cloud angle of that fight.

Practical takeaway for teams evaluating vendors now: map where agent telemetry is produced, which fields are sensitive, and which data must never leave your VPC or account boundary before you score feature checklists. Watch whether groundcover converts this round into deeper agent-specific signal coverage while holding the “never leave your cloud” constraint, and compare that posture against tools that assume SaaS-side aggregation. If your compliance story depends on residency, that architectural choice should sit above pricing and UI polish in the shortlist.