Adobe Puts an AI Assistant Inside After Effects — Beta Now, MCP Next

Adobe has announced an AI assistant for After Effects at Adobe Max 2026 that writes expressions — the snippets of JavaScript-style code motion designers use to automate animation — from plain-language prompts. The tool is available in beta now, with broader plumbing for third-party AI agents expected in early 2027.

Expressions are among the most powerful features in After Effects, and among the least approachable. They let editors and motion graphics artists build complex animations and automate repetitive tasks, but editors who don't write code routinely avoid them. Adobe's pitch is simple: describe the animation, and the assistant generates the expression.

At a briefing ahead of Max, the company demonstrated the assistant building an animation designed to resemble a collage moving at 18 frames per second. The editor described the intended result; the AI handled the code. No manual expression writing involved.

What does this change for working motion designers?

The generated expressions remain fully editable. That detail matters more than the demo itself — the output isn't a black-box render you accept or discard. You can fine-tune the animation, adjust values and intervene at the expression level, which is where client revisions actually happen.

Adobe also positions the assistant as a learning tool. Designers can examine the generated code to see how particular effects are achieved, which is a faster on-ramp to expressions than a manual page ever was.

The scope goes beyond animation code. According to Adobe, the assistant can also:

• organize complicated After Effects projects through conversational instructions
• build or debug existing expressions
• generate images and video from within the app

All of it runs on Adobe's Creative Agent, the company's cross-application AI framework being expanded across Creative Cloud.

Third-party agents get direct access

The more strategically interesting announcement is Adobe's plan to support the Model Context Protocol (MCP — an open standard that lets external AI applications talk to software directly). With MCP, users could connect their preferred third-party AI agents to After Effects rather than relying solely on Adobe's own tools.

Adobe says these agents will work with native project files locally on the user's computer, while keeping compositions and other elements editable. The MCP integrations are expected to arrive in early 2027. That's a genuine business-model shift for Adobe, which has historically kept its apps a closed shop, and it arrives as competing AI video and motion tools increasingly bypass traditional host applications altogether.

Shipped versus promised

Worth separating the two timelines here. The AI assistant itself exists today, in beta — demonstrable, with an editor watching a collage animation assemble from a sentence. The MCP support is a plan with a date attached, and plans with dates attached have a way of drifting.

There's also the standing industry argument about how much AI should sit inside the creative process at all. Fair debate. But expressions have always been a skill barrier that priced some editors out of automation entirely, and a tool that lowers it — while leaving the code visible and editable — doesn't remove craft. It just changes who gets to attempt it.

For studios weighing workflow investment, the practical question is whether beta access now is worth onboarding time before the MCP rollout firms up. Editors who already write expressions will likely use this as a debugging and scaffolding shortcut. Editors who don't may finally start using expressions at all.

The beta is live; the third-agent era arrives, by Adobe's own schedule, sometime next year.