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 default-deny Dogwood poli
What is 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 default-deny Dogwood policies and credential injection that keeps secrets outside the agent. Written in Rust. Supports macOS on Apple silicon and Linux.
Overview of box
What it is
Strands Box is an open source sandbox engine for AI agents. It combines operating-system isolation with semantic policies written in Dogwood, allowing control over what agents can access and the conditions under which they can act. Box is harness-agnostic: you choose the agent program or binary to run, configure its access in box.toml, and write its policies in policy.dw. The host OS enforces direct access restrictions, while Strands Shell, Monty for Python, the egress gateway, and the MCP broker send operations to the embedded Dogwood Local Engine for policy evaluation. Box currently supports local execution on macOS with Apple silicon; Linux support is planned.
Key features
- File, program, and network restrictions: the agent's sandbox limits direct access, and Box reports configured grants at startup.
- Semantic and temporal policies: Dogwood rules can depend on the requested operation, its arguments, earlier actions, and elapsed time. Operations checked by the policy engine are denied unless a rule permits them.
- Policy across code and tools: Strands Shell, Monty for Python, the egress gateway, and the MCP broker share one policy engine and event history, so a file read through one interpreter can affect whether a later network request is allowed.
- Request and tool-call checks: the egress gateway checks connections and HTTP requests (method and path); the MCP integration checks configured tool calls and their arguments.
- Credential injection: the gateway authenticates permitted requests with configured API credentials or AWS SigV4 signing; the agent does not receive the underlying secrets.
- Decision records: Box records policy decisions in OTLP JSON, by default to
/private/telemetry/records.jsonl.
Use cases
Running AI agents in a sandbox that restricts what they can execute, read, write, and reach on the network. Enforcing semantic and temporal policies across agent tools and interpreters. Injecting credentials for permitted network requests without exposing secrets to the agent.
Getting started
sh curl -fsSL https://raw.githubusercontent.com/strands-agents/box/main/download.sh | sh ./box-core/box --version
License and pricing
Apache-2.0, open source.
Based on the project's public README and repository metadata.