Sandbox MCP tools
Review the tools the Sandbox MCP gives your agent to create and work in a sandbox.
When you install the Sandbox MCP, your agent gains these tools for creating and working in a sandbox. Your agent chooses the tools as it works through your request. You do not call them yourself.
| Tool | What it does |
|---|---|
create_sandbox | Starts a sandbox and returns its ID. |
execute_code | Runs Python in the persistent kernel. |
execute_command | Runs a shell command, including elembio CLI commands. |
install_packages | Installs additional Python packages with pip. |
list_files | Lists the files in a sandbox directory. |
list_artifacts | Lists files with size and type details, to find a file to share. |
fetch_artifact | Creates a short-lived download link for a file in the sandbox. |
request_upload | Creates a short-lived link that uploads a file from your computer into the sandbox. |
get_status | Reports whether the sandbox is ready or busy, and how much memory it uses. |
get_results | Returns the result of code that continued running in the background. |
interrupt_execution | Stops the running code and keeps the kernel state. |
destroy_sandbox | Ends the sandbox permanently. |
read_skill | Reads an Element Biosciences skill that the Sandbox MCP provides, such as a multiomics skill. |