A dataset of sandboxes for running AI coding agents
please do not escape is a curated, evidence-carrying dataset of sandboxes and adjacent approaches for running AI coding agents with isolation, network control, or credential mediation. It solves the problem of finding reliable, secure sandbox solutions by providing a structured, filterable list where every entry links to primary sources. The dataset is for developers, teams, and researchers who need to evaluate sandbox environments for AI agents.
All data is stored in a structured YAML file, making it easy to parse and integrate.
Every claim in the dataset traces back to a primary source, ensuring accuracy and trustworthiness.
Users can hide quick picks or show non-product alternative approaches to narrow down options.
Anyone can submit a new entry or correct existing ones via pull requests on GitHub.
The dataset lives on GitHub at github.com/pjlsergeant/ai-sandboxes, enabling transparency and version control.
You can open a pull request on the GitHub repository or send additions via email to [email protected].
Yes, every claim in the dataset traces to a primary source, and contributions are encouraged to maintain accuracy.
The dataset covers sandboxes and adjacent approaches for AI coding agents, including isolation, network control, and credential mediation tools.
Yes, the dataset is publicly available on GitHub and the website can be browsed without any cost.