Dash0 has made Profiling in Dash0 generally available, adding continuous CPU and memory profiling to the observability platform. Announced during KubeCon + CloudNativeCon North America 2026 in Salt Lake City, the service is designed to show where processes spend compute resources and how they allocate memory, including at function and source-line level.
The profiler monitors processes across Kubernetes clusters and can also run on Linux hosts, virtual machines and bare-metal servers. Dash0 says it requires no code changes. It also accepts profiles submitted in the OpenTelemetry profiling format, providing a route for teams to send compatible profile data into the service.
The launch follows Dash0’s acquisition of Polar Signals, the Berlin-based company behind the open source Parca project, 12 weeks earlier. Dash0 is incorporating profiling as a native observability signal alongside traces, logs and metrics. The combination is intended to help engineers move from an alert or slow trace to the code paths consuming CPU or retaining memory.
Connecting profiles to traces
An eBPF agent runs on each node and profiles processes across a range of languages, including Go, Rust, C/C++, Java, Python, Node.js, Ruby, PHP and .NET. CPU profiles identify functions and lines associated with processor use. Memory profiles show call paths involved in allocations and which allocations remain live.
Dash0 links eBPF profiles to the traces they relate to. Engineers can therefore open a profile associated with a slow span and inspect which functions used the time represented by that span. The integration is aimed at making code-level evidence available in the same troubleshooting workflow as traces and logs, rather than requiring teams to treat profiles as a separate source of diagnostic information.
Dash0 says the profiler runs continuously on every node, so profile data is available from the start of an incident rather than beginning only after an engineer turns profiling on. The company cites measurements by Polar Signals indicating overhead below 1% even under heavy CPU load. That figure is the company’s reported measurement; the announcement did not provide details about the test configuration.
OpenTelemetry and access
The product also connects to Dash0’s work with the OpenTelemetry ecosystem. Polar Signals engineers contribute to the OpenTelemetry Profiling Working Group, where the profiling signal entered alpha in March 2026, and to the open source OpenTelemetry eBPF profiler. Parca remains open source following the acquisition.
For teams using coding agents, Dash0 says its Agent0 and agents such as Claude Code can read profiles through the Dash0 MCP server. This offers another way to make profiling information available to developer tools, while the core product provides the profile-to-trace connection for investigation.
Profiling is available to all Dash0 customers at a stated price of $0.60 per million samples. New users can access it through a 14-day free trial. The announcement positions continuous profiling as an always-on companion to existing telemetry: CPU and memory behavior can be examined alongside the traces, logs and metrics that first indicate a service problem.