BentoRun FAQs

Question 1

Why does BentoRun use NsJail instead of gVisor for sandboxing?

Accepted Answer

BentoRun chooses NsJail for its superior performance and lower overhead. NsJail's use of Linux namespaces is extremely lightweight, making it ideal for frequent, short-duration Python script executions typical of an MCP server, unlike the higher latency of nested containers like gVisor.

Question 2

What is BentoRun and its primary purpose?

Accepted Answer

BentoRun is a Model Context Protocol (MCP) server designed to provide AI agents with a secure, isolated environment for executing Python code, safeguarding the host system from untrusted or malicious scripts.

Question 3

Can Python code executed in BentoRun access external networks or APIs?

Accepted Answer

Yes, BentoRun sandboxes are configured to share the host container's network namespace, allowing executed Python code to make external API calls or install packages, a crucial feature for many AI agent tasks.

Question 4

What resource limits does BentoRun impose on executed code?

Accepted Answer

BentoRun enforces strict resource limits via cgroups and rlimits, including memory (default 512MB), CPU time (default 60s), and output size, preventing buggy or malicious code from exhausting system resources.

Question 5

How does BentoRun ensure code security and isolation?

Accepted Answer

BentoRun utilizes NsJail, a lightweight process isolation tool, leveraging Linux namespaces for PID, mount, and network isolation, cgroups for resource limits (memory, CPU), and seccomp-bpf syscall filters for precise control over allowed operations.

BentoRun

BentoRun

Key Features

Use Cases

Key Features

Use Cases