sandbox-runtime
by anthropics
A command-line tool and library that runs any program with strict limits on which files it can read or write and which websites it can reach. It uses the operating system's own sandboxing features, so no container is needed.
- Stars
- 6.9k
- Licence
- Apache-2.0
- Runs on
- Your computer
- Setup
- Moderate
- Language
- TypeScript
- Age
- 375 days
What you can do with it
- Run an MCP server inside a sandbox so it cannot write to sensitive folders
- Allow a command to reach only named websites and block all other network traffic
- Stop a script or AI agent from reading private files such as SSH keys
Make money with it
A freelancer could offer to set up secured AI agent workflows for clients, using sandbox rules as part of the service.
An idea, not financial advice. Check the licence before charging for anything built on it.
Ways a business could use it
- Limit the damage an AI coding agent or third-party tool can do on a work machine holding client data
- Safely try out unfamiliar scripts or MCP servers without giving them access to company files or the open internet
Can you use it commercially?
Free to use, change and sell, including in paid products. Keep the copyright notice.
Not legal advice. Read the project's licence file before building a product on it.
Getting it running
Install globally with npm (npm install -g @anthropic-ai/sandbox-runtime), then set rules in a JSON settings file; Linux needs bubblewrap installed.
Follow the official instructions in the project README.
Things to check
- Beta research preview: settings format and APIs may change
- Windows support relies on a dedicated user account and machine-wide firewall filters, with documented limitations
- Not a full container, so protection depends on correct configuration
Who's talking about it
- raiscui (gh-trends): GitHub trends — 2026-10-28 (daily release) Featured in a roundup
Checked 30 Oct 2026. Summary drafted by AI from the project's README.
