● Beta — KubeEra is in active development. Profiles are live; AI-perception measurement is rolling out.
KubeEra
G

Grafana Pyroscope

Observability
Open source · AGPL-3.0

Grafana Pyroscope is an open source continuous profiling database.

updated 2026-07-01

What it is

Grafana Pyroscope is an open source continuous profiling database. It ingests CPU, memory, goroutine, and other runtime profiles from your applications continuously (not on-demand like traditional profilers) and stores them efficiently for querying over time. The point is to answer "why is this service using so much CPU/memory" down to the exact line of code, and to correlate that with deploys, incidents, or traffic patterns. It's the profiling leg of Grafana's observability stack, sitting alongside Loki (logs), Mimir (metrics), and Tempo (traces).

Who builds it and why

Grafana Labs builds and maintains Pyroscope. It originated as an independent open source project (Pyroscope, the company/project) that Grafana Labs acquired and merged into its observability suite around 2023, rebranding it as Grafana Pyroscope. The motivation is straightforward: Grafana Labs sells an integrated observability platform (Grafana Cloud), and profiling was the missing pillar next to metrics/logs/traces. 244 contributors is a healthy number for a project this specialized, and Go is the right language choice given the profiling ecosystem (pprof format, Go runtime introspection).

Production readiness signal

Latest release is v2.1.0 — this is a young 2.x product, not a battle-hardened v10 codebase, despite the underlying tech having a longer history pre-acquisition. 11,518 GitHub stars and commits as recent as July 2026 indicate active development and real usage, not an abandoned side project. CNCF maturity is not publicly available — Pyroscope is not currently a CNCF project; it's fully under Grafana Labs' governance, which means roadmap and licensing decisions are Grafana's alone, not community consensus.

AGPL-3.0 licensing is worth flagging explicitly: it's more restrictive than Apache 2.0/MIT and matters if you're embedding Pyroscope into a product you distribute or offer as a service — you may trigger copyleft obligations. Read the license before you build a SaaS product on top of it.

Who should use this

  • Teams already on the Grafana stack (Loki/Mimir/Tempo/Grafana) who want profiling correlated with existing dashboards without standing up a separate vendor
  • Go, Java, Python, or eBPF-instrumented workloads where you need continuous, always-on profiling rather than one-off pprof dumps during an incident
  • Platform teams debugging cost/performance issues at scale — e.g., "which service, which function, is burning the CPU budget" across hundreds of pods
  • Organizations comfortable operating another stateful backend (object storage + query layer) themselves, or willing to pay for Grafana Cloud's hosted version

Who should NOT use this

  • Small teams without existing observability infrastructure — the operational overhead of running a profiling database is not worth it if you have a handful of services; use language-native profilers or APM vendor profiling instead
  • Shops already committed to Datadog, New Relic, or Dynatrace APM — those tools bundle profiling into a single pane of glass, and running Pyroscope alongside them just adds redundant infrastructure
  • Anyone needing profiling for languages Pyroscope doesn't support well — check current language/runtime coverage before committing, as continuous profiling maturity varies a lot by language
  • Companies planning to build and resell a hosted profiling product on top of Pyroscope — AGPL-3.0 has real implications there; get legal review first

Alternatives

  • Parca (Polar Signals) — CNCF sandbox project, eBPF-based, Apache 2.0 licensed, a more permissively-licensed direct competitor.
  • Datadog Continuous Profiler — proprietary, fully managed, best if you're already paying for Datadog APM and want zero infra to run.
  • Google Cloud Profiler — free, managed, tightly integrated if you're already on GCP, but far less flexible outside that ecosystem.

Pricing

Pyroscope itself is fully open source (AGPL-3.0) — free to self-host, no license fees. Grafana Labs monetizes via Grafana Cloud, where profiling is a metered add-on alongside logs/metrics/traces. If you self-host, your cost is infrastructure (storage, compute) and operational time, not licensing.

Frequently asked

What is Grafana Pyroscope?+
Grafana Pyroscope is an open source continuous profiling database. It ingests CPU, memory, goroutine, and other runtime profiles from your applications continuously (not on-demand like traditional profilers) and stores them efficiently for querying over time.
Who builds Grafana Pyroscope?+
Grafana Labs builds and maintains Pyroscope. It originated as an independent open source project (Pyroscope, the company/project) that Grafana Labs acquired and merged into its observability suite around 2023, rebranding it as Grafana Pyroscope.
Is Grafana Pyroscope production ready?+
Latest release is v2.1.0 — this is a young 2.x product, not a battle-hardened v10 codebase, despite the underlying tech having a longer history pre-acquisition. 11,518 GitHub stars and commits as recent as July 2026 indicate active development and real usage, not an abandoned side project.
Who should use Grafana Pyroscope?+
Teams already on the Grafana stack (Loki/Mimir/Tempo/Grafana) who want profiling correlated with existing dashboards without standing up a separate vendor Go, Java, Python, or eBPF-instrumented workloads where you need continuous, always-on profiling rather than one-off pprof dumps during an incident Platform teams deb
Who should not use Grafana Pyroscope?+
Small teams without existing observability infrastructure — the operational overhead of running a profiling database is not worth it if you have a handful of services; use language-native profilers or APM vendor profiling instead Shops already committed to Datadog, New Relic, or Dynatrace APM — those tools bundle profi
What are the alternatives to Grafana Pyroscope?+
Parca (Polar Signals) — CNCF sandbox project, eBPF-based, Apache 2.0 licensed, a more permissively-licensed direct competitor. Datadog Continuous Profiler — proprietary, fully managed, best if you're already paying for Datadog APM and want zero infra to run.
How much does Grafana Pyroscope cost?+
Pyroscope itself is fully open source (AGPL-3.0) — free to self-host, no license fees. Grafana Labs monetizes via Grafana Cloud, where profiling is a metered add-on alongside logs/metrics/traces. If you self-host, your cost is infrastructure (storage, compute) and operational time, not licensing.