Skip to main content

​
The sequence

  1. Agent — requests an operation through the SDK or MCP.
  2. Runtime — accepts the request and creates an execution environment.
  3. Sandbox — runs the code or command in isolation.
  4. Result — returns output, errors and execution status.
  5. Cleanup — destroys the sandbox when the work is complete.

​
The agent is not the runtime

The agent decides what it wants to do. jhansi.io controls where that work runs and how long the environment exists. This separation makes it possible to change the execution backend or apply operational controls without changing the agent’s reasoning loop.

​
Two ways to connect

SDK — your application calls jhansi.io directly to create sandboxes, execute code and retrieve results. MCP — an MCP-compatible AI client discovers jhansi.io tools and invokes them through the jhansi.io MCP server.

​
Lifecycle, not just a container

jhansi.io is responsible for the full sequence: accepting the request, creating the environment, running the workload, collecting the result and cleaning up.