Execute a command inside a running sandbox and return stdout/stderr/exit code.
AI agents invoke sandbox_exec to trigger actions in Microsandbox. What it does depends on the arguments the agent supplies, and its effects often reach beyond the immediate call — builds kicked off, notifications sent, workflows started.
This tool runs code/commands whose effects depend entirely on the arguments provided. While sandboxed (limiting blast radius to 'high' rather than 'critical'), an AI agent could still execute harmful commands, exfiltrate data, consume resources, or compromise the sandbox itself if given malicious instructions.
From the tool's definition Tool name 'sandbox_exec' combined with description 'Execute a command inside a running sandbox and return stdout/stderr/exit code' directly indicates execution of arbitrary commands.
Documented attack patterns abuse exactly the kind of access sandbox_exec gives an agent:
PolicyLayer is an MCP gateway — it sits between your AI agents and Microsandbox, and nothing reaches the server without passing your rules. This is the rule we recommend for sandbox_exec:
{
"version": "1",
"default": "deny",
"tools": {
"sandbox_exec": {
"limits": [
{
"counter": "sandbox_exec_rate",
"window": "minute",
"max": 10,
"scope": "grant"
}
]
}
}
}sandbox_exec stays usable, but rate-capped — a runaway agent can't fire it dozens of times a minute. Everything else on the server is denied unless you say otherwise.
Free to start. No card required.
Execute a command inside a running sandbox and return stdout/stderr/exit code. It is categorised as a Execute tool in the Microsandbox MCP Server, which means it can trigger actions or run processes. Use rate limits and argument validation.
Register the Microsandbox MCP server in PolicyLayer and add a rule for sandbox_exec: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Microsandbox. Nothing to install.
sandbox_exec is a Execute tool with high risk. Execute tools should be rate-limited and have argument validation enabled.
Yes. Add a rate_limit block to the sandbox_exec rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.
Set action: deny in the PolicyLayer policy for sandbox_exec. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.
sandbox_exec is provided by the Microsandbox MCP server (superradcompany/microsandbox-mcp). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Start from Microsandbox, add the rest of your stack, and see everything your agents can call. Then put policy on all of it.
Free to start. No card required.
19 Microsandbox tools catalogued and risk-classified — across an index of 43,000+ MCP servers.