Observability spent most of the past decade as a post-production discipline. It was, in many ways, a reactive practice: catching outages, reducing alert fatigue, and cutting the time engineers need to find a root cause. That focus is now shifting as agentic AI systems move into the software development lifecycle, pulling production context earlier into coding, testing, and deployment work.

This shift is helping to fuel growing demand for observability vendor Groundcover, which this week announced a $100 million Series C round. The company builds its observability technology on the open-source eBPF and OpenTelemetry technologies, and it has quickly become a notable player in the fast-evolving segment between traditional application performance monitoring and AI infrastructure observability.

From reactive monitoring to AI-native observability

Groundcover was founded in 2021 with a simple but ambitious goal: rethink how observability data is collected and processed. Rather than asking every developer to manually instrument each service with an SDK, Groundcover uses eBPF to capture telemetry from the Linux kernel itself. This approach removes a step that most observability vendors still require.

“You didn’t have to have the developer instrument an SDK, change their code base, and so on,” Azulay explained.

In its early days, that capability was mainly a convenience. It simplified the move from monolithic applications to microservices, and it helped teams get visibility across sprawling Kubernetes clusters without adding significant overhead. But today, the same property is becoming useful for a different reason: engineering teams are adopting new AI tools and coding agents so quickly that they often lose track of what is actually running in their own environment.

Azulay compared the current gap to the visibility problems teams dealt with roughly a decade ago, before observability tooling matured. eBPF operates below the application layer, rather than depending on code a developer wrote, so Groundcover can still see workflows that nobody thought to instrument.

“eBPF is kind of that security net of even if you didn’t instrument, even if you’re not in full control, you’re gonna know which agentic workflows are running in production, which models are using, which vendors they’re using, and so on,” Azulay said.

What eBPF does and why it matters more now

eBPF, short for extended Berkeley Packet Filter, is a Linux kernel technology that lets code run safely inside the kernel without a custom kernel module. It has long been used for network monitoring, packet filtering, and performance tracing. Groundcover uses eBPF to watch application and infrastructure activity without manual instrumentation, providing a unified source of truth across HTTP requests, database calls, queues, and system-level metrics.

The broader industry has grown increasingly interested in eBPF as an alternative to sidecar proxies and heavy agents. Because eBPF runs in the kernel, it can observe all system calls and network events without changing application code. That makes it especially attractive for AI workloads, where underlying dependencies change frequently and new models are introduced at a rapid pace.

OpenTelemetry, meanwhile, provides the standardization layer for exporting the data that eBPF captures. Together, they form the backbone of Groundcover’s platform. The company has also built its storage layer to sit inside the customer’s own cloud environment, rather than a shared vendor backend. That architectural decision, Azulay said, is built to hold the larger and more sensitive telemetry volumes that agentic workloads produce without shipping data to a third party.

“I think people are expected to save more telemetry, and save more telemetry more privately,” Azulay said.

How agentic workflows are breaking distributed tracing

Distributed tracing follows a request as it moves across services so engineers can see where time is spent and where something broke. It has always relied on a predictable number of hops, the kind of path an engineer could trace by hand, such as a cache calling a database. High-level visualization tools can present those traces in a straightforward waterfall graph.

Azulay said that assumption breaks down once agents enter the picture. A single agent session can generate a large number of tool calls and internal model calls, with no fixed pattern. The agent may decide to query a vector database, call a public API, invoke a code interpreter, and then call an LLM again, all within one user request. With LLMs and agentic workflows, he said, “this is becoming very complicated.”

The metrics that matter are also changing. Teams now track token usage, model costs, and hallucination rates alongside latency and error rate. Traces can contain a customer’s actual prompt instead of only structured request data, which raises privacy questions. As a result, Azulay said, AI observability is not exactly APM. “It’s not going to be the same product,” he said.

This distinction has major implications for the observability market. Traditional APM products were designed around bounded transactions with known service boundaries. AI observability, in contrast, must deal with open-ended processes that can last minutes or hours and exercise unpredictable combinations of resources. The data involved is also larger and less structured, making the choice of storage architecture critical.

Groundcover argues that its architecture is particularly well suited to this challenge. By keeping telemetry in the customer’s own environment, the platform can store more data without the cost or privacy concerns of shipping everything to a vendor-managed backend. It can also apply eBPF-based instrumentation to AI gateways, model APIs, and vector databases that do not always expose rich internal metrics.

Agent Mode and the rise of MCP

Groundcover is not just optimizing its platform for the needs of modern agentic AI activity. The company is also using AI to improve the user experience.

Agent Mode is Groundcover’s built-in AI assistant for engineers. It lets them ask questions about their systems, build dashboards, and troubleshoot problems in logs and traces without writing queries by hand. Instead of memorizing PromQL, LogQL, or proprietary query languages, engineers can interact with their telemetry through natural language.

Groundcover has also built a Model Context Protocol (MCP) integration that connects Agent Mode to coding agents and workflow tools including Linear, letting engineers and AI systems pass context back and forth during an incident. MCP is an open standard that gives AI models a structured way to call external tools, and it has quickly become a common integration layer for agentic development platforms.

Azulay said adoption of the MCP integration has moved faster than the company expected. Customers use the integration differently depending on how far along they are in adopting AI tools. Some ask questions through it instead of opening the Groundcover dashboard, while others use it to write a fix directly. He framed the pattern as an industry trend rather than a product win specific to Groundcover.

“People are basically starting to build their autonomous software development structure,” he said.

This shift toward autonomous software development is visible across the industry. AI coding assistants are no longer simple autocomplete tools; they can open issues, run tests, and even deploy code. For observability teams, that means the user of their tools is increasingly an AI agent, not just a human engineer. That creates new requirements for API access, machine-readable outputs, and fine-grained authorization.

Groundcover’s focus on MCP and natural-language interfaces appears to be an attempt to meet those requirements head-on. Instead of forcing agents to scrape dashboards or parse PDFs, the platform offers a direct protocol for retrieving telemetry and triggering actions.

Investor confidence and market context

The new $100 million Series C round is led by One Peak, with participation from Morgan Stanley Expansion Capital, Zeev Ventures, Angular Ventures, Heavybit, and Jibe. Including this round, Groundcover has raised a total of $160 million since its founding in 2021. The company is headquartered in Tel Aviv, Israel, and operates with a distributed engineering and go-to-market team.

The funding is a strong signal from investors who see observability as an essential layer for the AI era. The broader market has seen a wave of interest in AI infrastructure, but not all categories have benefited equally. Observability, in particular, has become one of the busiest spaces because every AI deployment still needs to be monitored, traced, and debugged. Yet the way that monitoring is done is changing, and Groundcover’s latest round suggests that investors are betting on eBPF-based approaches over manual instrumentation.

Groundcover also benefits from the open-source wave. OpenTelemetry has become the de facto standard for telemetry data, and eBPF is increasingly supported by major cloud providers and kernel distributions. By aligning with those standards, Groundcover can integrate into a wide range of environments without locking customers into a proprietary data format.

Groundcover at a glance

• Founded: 2021
• Total funding: $160 million
• Latest round: $100 million Series C, led by One Peak
• Other investors: Morgan Stanley Expansion Capital, Zeev Ventures, Angular Ventures, Heavybit, Jibe
• Headquarters: Tel Aviv, Israel
• CEO: Shahar Azulay
• What they do: Observability technology built on eBPF and OpenTelemetry

As agentic workflows continue to multiply, the distinction between monitoring and AI infrastructure will only become more important. Groundcover’s kernel-level visibility, private-by-design storage, and AI-native tooling position it to address a problem that most legacy observability vendors are not yet equipped to solve. The new capital will allow the company to expand its platform, deepen its agentic AI integrations, and bring its approach to a broader set of enterprise users.

Source: Network World News