What is AI2IN?

AI2IN gives your AI agents a secure, isolated Linux sandbox to run untrusted, model-generated code — hosted entirely in India (AWS Mumbai, ap-south-1).

Think of it as the runtime layer between your LLM and the real world: the model writes code, AI2IN runs it somewhere it can't hurt you, and hands back the output — stdout, charts, files, a live URL.

Why a sandbox?

AI-generated code is untrusted by definition. Running it on your own servers is a security and stability risk. AI2IN sandboxes isolate each execution with gVisor (a user-space kernel), so untrusted code runs against a syscall surface that never touches the host kernel.

What you can do

• Run code in a stateful code-interpreter kernel and get rich, typed results (text, charts, DataFrames).
• Pause and resume long-running agent sessions to save on idle compute.
• Run background processes (a dev server, a build) and stream their logs.
• Read and write files, and watch a directory for changes.
• Expose a port to a public HTTPS URL for a live preview.
• Provision managed databases (Postgres / Redis) with a connection string.
• Deploy persistent services from a Docker image to a live URL.

Why India?

Compute never leaves ap-south-1. For teams with data-residency requirements — regulated industries, government, anyone handling Indian user data — that's the point. Billing is in INR with GST.

Next steps

• Getting started — install and run your first sandbox.
• Core concepts — sandboxes, isolation, regions, auth.