> ## Documentation Index
> Fetch the complete documentation index at: https://docs.jhansi.io/llms.txt
> Use this file to discover all available pages before exploring further.

# The Agent Execution Problem

> Generated code becomes an infrastructure problem the moment an agent is allowed to run it.

An agent may need to inspect a repository, transform a file, query a service, run tests or call an internal API. Each task requires execution somewhere, with some level of access to data, networks and credentials.

## The unsafe shortcut

The easiest implementation is to let the agent invoke a local shell. That removes infrastructure work, but it also collapses the boundary between generated code and the user's machine.

## The externalised shortcut

A hosted sandbox moves execution away from the developer machine and can provide a mature isolated environment. But some teams cannot send their code or data to a third-party execution plane, or they need the sandbox to reach private systems from inside their own network.

## The jhansi.io approach

Keep the reasoning interface separate from the credential-bearing execution environment.

The agent sends an execution request. jhansi.io runs the work inside a sandbox in infrastructure chosen by the operator. The result comes back to the agent; the execution environment remains an operational concern.

## What jhansi.io optimises for

* Self-operated deployment.
* A small, understandable execution layer.
* SDK and MCP access.
* Explicit sandbox lifecycle.
* A future path to credential and governance controls.