> ## 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.

# MCP Server

> Connect AI coding agents to jhansi.io sandboxes via the Model Context Protocol.

The jhansi SDK includes a built-in MCP server, letting AI agents (Claude Code, Cursor, Windsurf) create and run sandboxes directly.

## Start the server

```bash theme={null}
python -m jhansi.mcp_server
```

Configure the endpoint via environment variable:

```bash theme={null}
export JHANSI_BASE_URL=http://localhost:8000
```

## Claude Code

```bash theme={null}
claude mcp add jhansi --scope user \
  --env JHANSI_BASE_URL=http://localhost:8000 \
  -- python -m jhansi.mcp_server
```

## Tools

| Tool             | Description                                            |
| ---------------- | ------------------------------------------------------ |
| `create_sandbox` | Create a new isolated sandbox. Returns the sandbox ID. |
| `exec_code`      | Execute a command in a sandbox. Returns the output.    |
| `delete_sandbox` | Delete a sandbox and free its resources.               |