Overview

Experiments are declared in JSON or YAML, which lets teams keep them alongside other code and use familiar collaboration and orchestration practices. Each experiment sets a steady-state hypothesis, uses probes to validate it, and defines a method of actions and probes. If steady state is not met, the method is withheld and the experiment must bail out. That is a useful safeguard for teams injecting faults, though it does not replace careful experiment design.

The offering is a standalone CLI, not the separate Reliably platform. Teams choosing Chaos Toolkit take responsibility for installing and operating it in their own environment rather than selecting Reliably’s SaaS, private SaaS or on-premises options.

Key features

Experiments as code

The CLI can be embedded in a CI/CD chain, making it a fit for teams that want experiments in their delivery workflow. Its Open API supports extensions for different systems. After installing the CLI and core library, users can add extensions for the facets of chaos engineering they need. The extension directory includes AWS, Azure, Google Cloud Platform, Kubernetes, Kafka, Slack, Datadog, Grafana, OpenTelemetry and Prometheus.

Fault injection and controls

Chaos Toolkit supports Kubernetes, cloud and network fault injection, experiment scheduling and blast-radius controls. That combination gives teams several ways to scope and automate experiments, rather than limiting them to writing definitions. The toolkit’s extension model is also relevant to teams working across different systems, but using it means managing the CLI and extensions themselves.

Open source and community

Apache License 2.0 and an invitation to contribute through issue reports, voting, stars, Slack, pull requests and extension proposals make the project suitable for teams that value an open development model. Its contribution guide recommends enabling two-factor authentication on contributor accounts as a measure against account compromise.

Pricing

Chaos Toolkit — free. The plan is an open-source CLI, free forever under Apache License 2.0. There is no paid plan in this offering; it is a strong fit for teams that want to run the toolkit themselves without a software charge, but it is not a substitute for a managed service.

Platforms

The deployment model is self-hosted, with platform coverage listed for API, Linux, macOS, web and Windows. The CLI is implemented in Python 3, officially supports Python 3.8 and later, and has been tested only against CPython. Installation instructions cover Python setup on macOS, Debian or Ubuntu, and Windows, followed by installation of the CLI with pip. That gives teams several supported setup paths, while making Python runtime management part of adoption.

Who it's for

Choose Chaos Toolkit if your team wants code-defined experiments, a safety condition before fault-injection methods run, and the ability to add system integrations or place experiments in CI/CD. It is less suitable if you want a managed service, or if your team does not want to operate a Python CLI and its extensions.

Pros and cons

Useful precondition for experiments: a failed steady-state check prevents the method from running, reducing the chance of proceeding when the declared hypothesis is not satisfied.
Works with code-based workflows: JSON or YAML experiment files and CI/CD embedding suit teams that want experiments treated like other code.
Extensible across systems: the Open API and integrations for cloud, orchestration, messaging and observability systems give teams options to adapt the toolkit to their environment.
Free and open source: the CLI is free and uses Apache License 2.0.
Requires self-management: users install the Python CLI and extensions themselves; teams looking for an operated platform should consider a separate offering.
Python constraint: the CLI officially supports Python 3.8+ and has only been tested against CPython, which may narrow runtime choices.

Alternatives

Chaos Engineering Platforms is the broader category to browse when comparing options.

Reliably is worth considering if you want a separate platform and a free plan capped at one team member; it describes itself as open source and says its approach avoids lock-in.

ChaosBlade is another free, open-source chaos engineering tool, with support for three system platforms, four programming-language applications and more than 200 experiment scenarios.

FaultMesh may suit teams seeking a hosted option with a free Starter plan for up to three services, core fault types, basic resilience scoring and community support, or a Growth plan priced at 49.00 USD per month.

Krkn is another free option if you want a tool with no stated pricing or usage limits.

Toxiproxy is a free, open-source TCP proxy under the MIT license, an alternative for readers looking specifically for that kind of tool.

Chaos Mesh is a free, open-source, cloud-native chaos engineering platform that runs on Kubernetes, making it a natural alternative for Kubernetes-focused teams.

Gremlin is an option if you prefer a paid platform and want to evaluate it through its 30-day free trial, which requires no credit card and does not auto-convert.

Steadybit is another paid option; its Enterprise plan includes on-prem installation, SAML, audit log and enterprise support.

Verdict

Chaos Toolkit is a good choice for developers who want a free, code-oriented way to run scheduled chaos experiments with a steady-state gate and self-hosted control. Choose it for its open, extensible CLI and workflow fit; look elsewhere if you want a managed platform rather than a Python tool your team operates.

	Chaos Mesh	Chaos Toolkit
Free plan	Yes	Yes
Paid from	Free	Free
Platforms	api, self-hosted, Web	api, Linux, macOS, self-hosted, Web, Windows
Free plan	Yes	Yes
Kubernetes support	Yes	Yes
Cloud fault injection	Yes	Yes
Network fault injection	Yes	Yes
Experiment scheduling	Yes	Yes
Deployment model	self_hosted	self_hosted
Blast-radius controls	Yes	Yes

Chaos Mesh vs Chaos Toolkit

Chaos Mesh

Chaos Toolkit