A new project called aftr has been released, allowing Adobe After Effects to be controlled by AI agents such as Claude Code and Cursor. The tool acts like Puppeteer for video editors, providing a programmatic interface to automate complex tasks via WebSocket and the Model Context Protocol (MCP).
What Happened
A developer has introduced aftr—a system that transforms Adobe After Effects into a programmatically controllable tool. The project utilizes MCP and WebSocket to facilitate interaction between high-level LLM agents and the software's low-level API. The system includes a built-in autonomous orchestrator capable of independently creating video segments, performing rendering, and conducting visual verification of the results.
Context
Traditionally, After Effects is a heavyweight software with a graphical user interface (GUI) that requires manual management of layers and effects. Using the Model Context Protocol (MCP) allows for the standardization of interaction between AI agents and professional software, creating a bridge between text commands and complex video editing logic.
Why It Matters for the Industry
The project demonstrates a shift toward the Autonomous Video Production model. This paves the way for fully autonomous video production houses, where the industry can move from manual editor control to declarative scene descriptions via AI. The emergence of specialized agentic wrappers for media software could lead to the formation of a new niche: AI-native motion design.
Why It Matters for Users
For motion designers and users, this means the ability to automate routine tasks such as layer management and rendering, as well as quickly prototype complex video effects using simple text commands. This significantly accelerates workflows, allowing professional-level tasks to be assigned without manual interface manipulation.
Uncertainties / Limitations
There are legal risks associated with potential violations of Adobe's Terms of Service (ToS), as well as uncertainty regarding the legal status of AI as an autonomous content creator.
Sources
Author
Look at AI, Editorial Team