CodSpeed
Continuous performance testing platform that catches regressions in pull requests using CPU simulation, differential flamegraphs, and automated benchmarking for Python, Rust, Node.js, Go, and C/C++.
At a Glance
About CodSpeed
CodSpeed is a continuous performance testing and optimization platform that integrates directly into CI/CD workflows to catch performance regressions before they ship. It supports multiple languages and benchmark frameworks, delivers results as PR comments and status checks, and uses CPU simulation to bring measurement variance below 1%. The core CLI is open source under the Apache License 2.0, while the cloud dashboard and hosted runner infrastructure are commercial offerings.
What It Is
CodSpeed sits between your benchmark suite and your pull request workflow. Developers write benchmarks using familiar frameworks — pytest, criterion, vitest, divan, google-benchmark, or Go's built-in testing package — and CodSpeed wraps those runs with instrumentation that filters out system noise and produces stable, comparable results. The platform then surfaces performance changes as PR comments, blocks merges that exceed configured thresholds, and provides differential flamegraphs to pinpoint which lines of code got slower.
How the Instrumentation Works
CodSpeed offers three measurement modes, each suited to different workloads:
- CPU Simulation: Simulates CPU behavior to achieve less than 1% variance regardless of system load. Hardware-agnostic and automatically generates flamegraphs. Best for CPU-intensive code and CI regression detection.
- Walltime: Measures real-world execution time including I/O, system calls, and multi-threading effects. Recommended for API tests and I/O-heavy workloads, and works best on CodSpeed's hosted Macro Runners rather than standard cloud VMs.
- Memory: Tracks heap allocations using eBPF profiling, covering peak usage, allocation count, and allocation size. Supported for Rust and C/C++ with libc, jemalloc, or mimalloc.
Setup Path
CodSpeed provides two setup routes. The AI Wizard analyzes a connected repository, detects or creates benchmark files, generates a CI workflow configuration, and opens a pull request — requiring no manual configuration. The manual path involves installing a language-specific harness (e.g., pip install pytest-codspeed), writing benchmarks with the appropriate marker or framework, and adding a GitHub Actions or GitLab CI workflow that calls the CodSpeedHQ/action. The CLI can be installed via a single shell command and officially supports Ubuntu 20.04/22.04/24.04 and Debian 11/12.
Language and CI Integrations
CodSpeed provides first-class harness integrations across:
- Python — pytest plugin (
pytest-codspeed) - Rust — divan, criterion.rs, bencher
- Node.js — vitest, tinybench, benchmark.js
- C/C++ — google-benchmark
- Go — built-in
testingpackage - Zig — community-maintained harness
CI integrations cover GitHub Actions (via the official CodSpeedHQ/action), GitLab CI, and Buildkite. The platform also supports a codspeed.yml configuration file for defining multiple benchmark targets with per-benchmark options like warmup time and max rounds.
Update: v4.18.1
The latest CLI release is v4.18.1, published on June 24, 2026. The GitHub repository, written primarily in Rust, was last pushed on June 30, 2026, indicating active development. The project has accumulated over 210 stars and 24 forks since its creation in November 2023. Recent documentation highlights the AI Wizard as a new automated setup path, and the platform page lists Go support as a current language integration alongside the established Python, Rust, Node.js, and C/C++ support.
Community Discussions
Be the first to start a conversation about CodSpeed
Share your experience with CodSpeed, ask questions, or help others learn from your insights.
Pricing
Free
For small teams and open-source projects
- Unlimited runs & repositories
- Up to 5 users (unlimited for open-source projects)
- 3-month history
- CI integration
- Pull-Request reports & checks
Pro Trial
14-day free trial of the Pro plan, no credit card required
- All Pro features
- Unlimited history
- Priority support
Pro
For teams scaling automated performance testing
- All Free features
- Seat-based pricing
- Unlimited history
- Priority support
- Macro Runners usage add-on available
- Database instruments add-on available
- GPU instruments add-on available
Enterprise
For large-scale performance optimization and maximum security
- All Pro features
- SSO/SAML
- Role management
- SOC2 Type 2 reports
- On-premise deployment
- Custom Macro Runners (GPU, OS, architecture)
- Performance Data API
- Dedicated support with SLAs
- Invoiced billing
- Dedicated GitHub App
Capabilities
Key Features
- Continuous performance regression testing in pull requests
- CPU simulation instrument with <1% measurement variance
- Walltime instrument for I/O and multi-threaded workloads
- Memory instrument with eBPF heap allocation profiling
- Differential flamegraphs to pinpoint performance regressions
- PR comments and status checks with performance impact
- Merge protection to block PRs that degrade performance
- AI Wizard for automated benchmark setup and CI configuration
- Multi-language support: Python, Rust, Node.js, Go, C/C++, Zig
- GitHub Actions, GitLab CI, and Buildkite integrations
- Hosted Macro Runners (bare-metal ARM64) for consistent walltime measurement
- Performance gating with customizable thresholds
- Local benchmark runs via CLI
- codspeed.yml configuration for multi-benchmark definitions
- Differential profiling history and performance trend tracking
