Skip to main content
The agent may be an AI coding assistant, an automated workflow, a custom application using an LLM, or an MCP-compatible client. It decides that a task should run and supplies the necessary input.

​
What the agent sends

  • The requested operation.
  • Code, commands or input data.
  • Execution options exposed by jhansi.io.

​
What the agent receives

  • Execution status.
  • Standard output and error.
  • Structured results or explicitly returned files.

​
What the agent should not own

The agent should not need direct access to the host, runtime internals or a general credential store. jhansi.io exists to preserve that separation.