AI agents can now run Kling through MCP or the CLI

A remote MCP server and a CLI let AI assistants and scripts generate images and videos with Kling, using paid credits from a personal workspace.

A compatible AI assistant can request an image or video from Kling without using its web-based creation interface. Two options provide access to its generation tools: a remote MCP (Model Context Protocol) server and a command-line interface, or CLI.

The MCP connection is configured directly in a compatible assistant’s settings, such as Claude or Cursor. For international accounts, the server URL is `https://kling.ai/mcp`. After completing OAuth authorization, users can describe their request in natural language and receive the result in the assistant. No separate Skill installation is required for this connection.

The CLI provides access to generation capabilities through a terminal, scripts, or automated workflows. For use through an AI agent, the official Skill can be installed with `npx skills add klingai-tech/skills`. It gives the assistant instructions for using the CLI. The npm packages are distributed separately: `@klingai/cli-global` for international accounts and `@klingai/cli-cn` for Chinese accounts. The `kling login` command opens browser-based OAuth authentication. This Skill supports CLI use; it is not required for a direct MCP connection.

The official guide recommends choosing one integration per assistant, whether plugin, MCP, or CLI, to prevent unpredictable switching between multiple options.

Only paid credits from a personal workspace are supported. Bonus credits, free off-peak generations, and team workspace benefits are excluded. The rate limit is five requests per second, with one concurrent video generation task for non-subscribers. Result URLs remain valid for 24 hours; submitted tasks cannot be canceled.