Polar Signals

Polar Signals develops continuous profiling technology for applications and infrastructure. Its commercial product, Polar Signals Cloud, is a hosted service that stores and analyzes profiling data collected by a zero-instrumentation eBPF agent deployed on customer nodes. The agent runs on Kubernetes, Docker, ECS and bare metal, requires no code changes, and is described by the company as adding under 1% overhead, making it suitable for continuous use in production. Supported languages include C/C++, Rust, Go, Ruby, Python, PHP, JVM languages (Java, Clojure, Kotlin), Graal Native, Node.js and .NET.

The platform captures CPU, memory and GPU data. For NVIDIA GPUs it uses CUPTI together with eBPF to provide per-kernel timing, CPU stack traces and CUDA Graph support without filesystem writes or network overhead, positioning it against traditional CUDA profilers that are designed for debugging rather than live systems. Analysis features include flame-graph visualization, function-level filtering, Prometheus-style label selectors and a query language across services and environments, side-by-side comparison of two profiling queries to identify regressions, automated detection of regressions and optimizations over time down to the code line, and OpenTelemetry compatibility with an optional optimized collection protocol. An MCP integration exposes profiling data to LLM-based tools such as Claude Code, Cursor, Copilot and Gemini so they can analyze performance data and suggest source-code changes. Product administration features include organizations and projects, user invitations and roles, single sign-on, token generation and debuginfo uploads.

Polar Signals is the primary developer of Parca, an open-source continuous profiling project that tracks CPU and memory usage over time and forms the basis of the commercial offering. The company also built "Great Lakes," a purpose-built storage engine for profiling data and the successor to FrostDB.

Founding story

Polar Signals was founded in 2020 by Frederic Branczyk, a former Red Hat engineer and prominent contributor in the Prometheus and Kubernetes communities. The company was founded in Berlin, and Branczyk's premise was that continuous, accessible performance data is necessary in modern cloud-native software development and operations. The team was assembled from engineers previously at Red Hat, CoreOS and the Prometheus and Kubernetes maintainer communities, and its early work centered on the open-source Parca project, which became the basis for the commercial cloud product.

Business model

Commercial SaaS built on an open-source core. Polar Signals maintains the open-source Parca project and sells Polar Signals Cloud, a hosted continuous profiling service that adds enterprise features such as single sign-on, team provisioning and permission management. The site references usage/volume-based pricing questions (monthly billing basis, volume discounts for high-volume data profiling), a free 14-day trial, and procurement through Google Cloud Marketplace.

Subscription to Polar Signals Cloud with monthly billing; FAQ topics cover how pricing is determined, the basis for monthly billing, additional fees and discounts for high-volume profiling data. A free 14-day trial is offered and the product can also be procured via Google Cloud Marketplace.

Traction

By March 2024 the company reported more than a dozen paying customers following its commercial launch the previous October, including Vercel, Materialize, Canonical and Weaviate, and said its pipeline exceeded its ability to close deals. Public testimonials cite users at Indent, Zalando, Authzed and Thought Machine, including a Thought Machine SRE reporting the discovery of a workload consuming 55% of CPU within 15 minutes of use. The company had 11 employees at that time. It holds AICPA SOC audit coverage and is available through Google Cloud Marketplace.

Latest developments

On 17 August 2026 Dash0 announced its acquisition of Polar Signals, with terms undisclosed. Dash0 plans to embed continuous profiling natively in its SignalStore backend and Agent0 product, use Polar Signals' Great Lakes storage engine to replace ClickHouse in SignalStore, and introduce related capabilities including AutoTune (automated pull requests to optimize CPU and memory usage) and SignalControl (dynamic adjustment of profiling granularity during anomalies). Frederic Branczyk framed the deal as extending the reach of the company's continuous and GPU/CUDA profiling work and connecting it to AI agent workflows. Prior to the acquisition, the company had shipped production-safe GPU profiling, MCP integrations with LLM coding tools, NVIDIA Inception Program participation and a Google Cloud Marketplace listing.