strands-agents/box: Run AI agents in a sandbox that restricts what they can execute, read, write, and reach on the network. Box combines OS isolation with defau
Run AI agents in a sandbox that restricts what they can execute, read, write, and reach on the network.
ProofCode ↗
Key points
- Box combines OS isolation with default-deny Dogwood policies and credential injection that keeps secrets outside the agent.
- Supports macOS on Apple silicon, with Linux support planned.
- Topics: agentic-ai, agents, ai, ai-agents, ai-safety, containment, egress, harness, linux, llm, macos, mcp, policy, rust, sandbox, sandboxing, seatbelt, security, strands-agents, zero-trust.
Sources (1)
- [1]strands-agents/box: Run AI agents in a sandbox that restricts what they can execute, read, write, and reach on the network. Box combines OS isolation with defauRising AI repositories on GitHub · Oct 2, 08:34 PM
Run AI agents in a sandbox that restricts what they can execute, read, write, and reach on the network.
Box combines OS isolation with default-deny Dogwood policies and credential injection that keeps secrets outside the agent.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 2, 2026Anthropic invests $100 million to train 10,000 engineers and tackle the enterprise AI talent gap
- Oct 1, 2026Claude-shaped science
- Sep 30, 2026Last Week in AI #345 - 5 new models, 9 misalignment incidents, some Dots
- Sep 2, 2026Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
- Aug 13, 2026Introducing Gemini 3.7 Flash
- Jul 21, 2026Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber
