AutoDL

•
4
•

Programmatically manage AutoDL GPU instances from the command line, an MCP server, or an SDK, enabling creation, boot, SSH, execution, and automated shutdown.

AutoDL-cli provides a versatile suite of tools for programmatically managing AutoDL GPU instances, addressing the challenge of AI agents needing on-demand GPU resources. It offers a command-line interface for human users, an MCP server for integration with coding agents, and an SDK for Node.js applications, all built on AutoDL's official open API for stability. This empowers users and agents to create, configure, run tasks on, and shut down GPU instances efficiently, complete with cost-saving features like automatic shutdown and detailed usage monitoring, ensuring that expensive GPU resources are utilized judiciously.

主な機能

01Automated GPU instance creation, power management, SSH, and file transfers.
02Cost-guard features including TTL auto-shutdown, idle detection, and balance gate.
030 GitHub stars
04Agent-friendly CLI, MCP server, and Node.js SDK for seamless integration.
05Interactive TUI dashboard for real-time monitoring of instances, costs, and TTL.
06Single-command `run` and `deploy` workflows for end-to-end task execution and project deployment.

ユースケース

01Allowing AI agents (e.g., Claude Code, Cursor) to provision and manage GPU instances dynamically mid-task.
02Automating the training and deployment of machine learning models on AutoDL GPU instances.
03Monitoring and controlling GPU instance costs through automated shutdown mechanisms and a live dashboard.