Chaos Toolkit
chaostoolkit.org
- Free plan
- Open source
- Windows
- macOS
- Linux
- Web
- API
- Self-hosted
What Chaos Toolkit is
Chaos Toolkit is a free, open-source toolkit for developers who want to find weaknesses in systems through chaos engineering. An experiment sets out a steady-state hypothesis and uses probes to check it, then defines actions and probes for its method. If the system fails the steady-state check, the method is not run and the experiment must stop. Experiments can be saved in JSON or YAML, making them usable in collaboration and orchestration alongside code. The toolkit has an Open API for extending support to systems and can be embedded in a CI/CD chain. Listed integrations include AWS, Azure, Google Cloud Platform, Kubernetes, Kafka, Slack, Datadog, Grafana, OpenTelemetry and Prometheus. The standalone CLI is self-hosted, free forever, and licensed under Apache License 2.0. It runs on Python 3.8 or later and has only been tested with CPython. Installation instructions cover macOS, Debian or Ubuntu, and Windows; extensions can be added after installing the CLI and core library.
Who it suits
It suits developers who want to run controlled chaos engineering experiments and integrate them into a CI/CD chain. Teams seeking a self-hosted, extensible CLI may find its JSON or YAML experiment format useful.
Good to know, and what to watch
- Free, open-source CLI under Apache License 2.0
- Experiments can be stored as JSON or YAML
- Embeds in a CI/CD chain
- Listed integrations include Kubernetes and major cloud platforms
- Safety gate stops methods when steady state fails
- CLI officially supports Python 3.8 and later
- Tested only against CPython
Plans
Every plan and priceChaos Toolkit
Free plan
Getting started
- Visit the Chaos Toolkit website
- Install Python 3.8 or later on macOS, Debian or Ubuntu, or Windows
- Install the CLI and core library with pip
- Install extensions for the systems or chaos facets you need
- Create an experiment file in JSON or YAML
- Run the experiment from your CLI or embed it in a CI/CD chain
The Club's review
Chaos Toolkit is an open-source toolkit for designing and running chaos experiments against systems. It is a strong fit for teams that want experiments alongside their code and can manage a Python CLI; its safety gate is valuable, but the toolkit is self-hosted rather than a managed service.
Overview
Experiments are stored as JSON or YAML, which lets teams collaborate on them and orchestrate them like other code. Each experiment sets a steady-state hypothesis, uses probes to check it, then defines a method of actions and probes. If the system does not meet the steady-state condition, the method is withheld and the experiment must stop. That is a useful guardrail before injecting faults, though it does not replace careful planning of what an experiment may affect.
The project is licensed under Apache 2.0 and offered as a standalone open-source CLI. That keeps the toolkit free to run under a self-hosted deployment model, but teams must manage the runtime and environment themselves. Reliably is a separate offering, with SaaS, private SaaS and on-premises options.
Key features
Experiments, automation and integrations
Chaos Toolkit can be embedded in a CI/CD chain, so teams can make experiments part of a delivery workflow. Its Open API allows extension for different systems; after installing the CLI and core library, users can add extensions for particular facets of chaos engineering. Integrations include AWS, Azure, Google Cloud Platform, Kubernetes, Kafka, Slack, Datadog, Grafana, OpenTelemetry and Prometheus.
The toolkit supports Kubernetes, cloud fault injection, network fault injection, experiment scheduling and blast-radius controls. That combination gives teams a broad set of ways to exercise system behavior, while the extension model means setup depends on installing the relevant components rather than expecting every integration in the core CLI.
Python CLI and setup
The CLI is implemented in Python 3, officially supports Python 3.8 and later, and has only been tested against CPython. Installation uses pip, with setup guidance for macOS, Debian or Ubuntu, and Windows. This is a reasonable route for teams already equipped to manage Python tooling; it is a meaningful constraint for organizations that want a packaged, centrally operated service instead.
The project invites issue reports, issue voting, stars, community Slack participation, pull requests and extension proposals. Its contribution guide also recommends enabling two-factor authentication on contributor accounts to reduce risk if an account is compromised.
Pricing
Chaos Toolkit — free. The plan is the open-source CLI, free forever under the Apache License 2.0. There is no paid tier in this standalone toolkit plan; users install the CLI and core library, then add extensions as needed. Its cost advantage comes with self-hosted operation and Python-based setup rather than managed service convenience.
Platforms
Chaos Toolkit is available for API, Linux, macOS, self-hosted, web and Windows environments. Its installation guidance covers macOS, Debian or Ubuntu, and Windows, while the CLI's Python 3.8+ requirement and CPython-only testing should guide compatibility planning.
Who it's for
Choose Chaos Toolkit if your team wants versionable experiment definitions, CI/CD integration and a free, extensible toolkit it can operate itself. It particularly suits developers and reliability teams comfortable with Python and with selecting and installing extensions for their target systems. Look elsewhere if you need a managed deployment, or if maintaining a Python CLI is not a fit for your team.
Pros and cons
- Pros: JSON or YAML experiments can live and evolve alongside application code, supporting collaboration and orchestration.
- Pros: The steady-state check blocks the experiment method when its precondition fails, providing a concrete safety gate.
- Pros: The free-forever Apache-licensed CLI supports CI/CD use and can be extended for a range of cloud, infrastructure, observability and messaging systems.
- Cons: The offering is a self-hosted CLI, not a managed service; teams must own its installation and operating environment.
- Cons: Python 3.8+ and CPython-only testing narrow the supported runtime expectations compared with tools that do not require this CLI setup.
- Cons: Extensions are installed separately after the CLI and core library, adding a setup step for teams with specific integration needs.
Alternatives
For a broader comparison, browse Chaos Engineering Platforms.
Consider FaultMesh if you want a freemium option with a free Starter plan for up to 3 services and a paid Growth plan at $49 a month, billed per h.
Reliably is a separate option if you want its stated open-source approach and a choice of SaaS, private SaaS or on-premises deployment.
ChaosBlade is another free open-source chaos engineering tool, described as supporting 3 system platforms, 4 programming-language applications and 200+ experiment scenarios.
Krkn is another free chaos engineering option.
Choose Chaos Mesh if you want a free, open-source chaos engineering platform that runs on Kubernetes.
LitmusChaos offers both a free hosted SaaS plan and a free open-source platform for installation on your own cluster.
Toxiproxy is a free, open-source TCP proxy under the MIT license.
Pumba is a free, open-source container chaos testing tool under Apache License 2.0.
Verdict
Chaos Toolkit is a compelling choice for developers and reliability teams who want free, code-oriented experiments with a clear steady-state gate and the flexibility to extend their setup. Its self-hosted Python CLI is also its main trade-off: teams that want managed operations or less runtime ownership should choose another platform.
The verdict
Chaos Toolkit gives developers a code-oriented way to define experiments, check steady state, and halt when that check fails. Its free CLI, extensions and integrations make it suited to teams comfortable with Python-based setup.
Questions people ask
Does Chaos Toolkit have a free plan?
Yes. The Chaos Toolkit plan costs nothing, is free forever, and is released under Apache License 2.0.
Which operating systems are covered?
Installation instructions cover macOS, Debian or Ubuntu, and Windows. The toolkit is also described for API, Linux, macOS, self-hosted, web and Windows environments.
What Python versions are supported?
The CLI is implemented in Python 3 and officially supports Python 3.8 and later. It has only been tested against CPython.
Can experiments be integrated with CI/CD?
Yes. Chaos Toolkit can be embedded in a CI/CD chain, and experiments can be stored as JSON or YAML files.
Which integrations are listed?
The extension directory lists AWS, Azure, Google Cloud Platform, Kubernetes, Kafka, Slack, Datadog, Grafana, OpenTelemetry and Prometheus.
Is Chaos Toolkit the same as Reliably?
No. Chaos Toolkit is presented as a standalone open-source CLI. Reliably is described separately with SaaS, private SaaS and on-premises options.
What the maker publishes
- Automation
- The toolkit can be embedded in a CI/CD chain. source
- Automation
- The toolkit is designed to run inside CI/CD workflows. source
- Community
- The project invites contributions through issue reports, issue voting, project stars, community Slack, pull requests and extension proposals. source
- Community support
- The project directs users to community Slack, Stack Overflow, GitHub issues and a project email address. source
- Container architectures
- Docker Hub lists official images for linux/amd64 and linux/arm64. source
- Containers
- Official container documentation lists latest, basic and full images, with Alpine and Debian-based variants. source
- Core projects
- The project includes separate CLI, core-library, Kubernetes, addons, reporting and bundler projects. source
- Deployment options
- The offering page describes a standalone open source CLI; it distinguishes Reliably as a separate platform with SaaS, private SaaS and on-premises options. source
- Discovery
- The chaos discover command can inspect an integration and produce a discovery report used to bootstrap experiments. source
- Experiment as code
- Experiments can be stored as JSON or YAML files so teams can review, version and orchestrate them like code. source
- Experiment execution
- A basic experiment is run with the command chaos run experiment.json. source
- Experiment format
- Experiments can be declared and stored as JSON or YAML files for collaboration and orchestration like other code. source
- Experiment model
- Experiments use probes to collect system information and actions to perform operations against the system. source
- Experiments
- An experiment declares a steady-state hypothesis and probes to validate it, plus a method made up of actions and probes. source
- Extension methods
- Extensions can be implemented as Python functions, external processes or HTTP endpoints. source
- Extensions
- After installing the CLI and core library, users can install extensions for different facets of chaos engineering. source
- Independent limit: extension dependency
- Hysen Labs says platform-specific actions depend on separately installed extensions and that extension maintenance can become a compatibility risk. source
- Independent limit: hosted control plane
- Zaira Labs says the file-driven CLI lacks the centralized scheduling, multi-tenant isolation and web UI associated with hosted chaos platforms. source
- Independent limit: interface
- Gremlin reports that the toolkit is CLI-only, which makes multi-system execution difficult without suitable cloud or orchestration drivers. source
- Independent limit: missing features
- Gremlin’s comparison lists no native scheduling, GUI or REST API in the core toolkit. source
- Independent praise: control
- Gremlin highlights the toolkit’s user-defined experiments and native rollback capability as strengths. source
- Independent praise: portability
- Gremlin’s comparison says Chaos Toolkit lets users define their own experiments and supports Docker, Kubernetes, bare-metal and cloud platforms. source
- Installation
- The installation guide provides Python setup instructions for macOS, Debian or Ubuntu, and Windows, and installs the CLI using pip. source
- Installation
- The documented installation method is pip install -U chaostoolkit inside a Python environment. source
Checked against the maker's pages on 15 Oct 2026.