After Effects has never been limited to what you can click in its interface. Expressions can connect properties, JSX scripts can automate project operations, panels can package repeatable tools, and aerender can send an approved composition through a controlled render pipeline. Claude Code makes those existing automation surfaces much easier to approach.

Instead of manually translating every request into expressions, scripts, layer operations, and output settings, a motion designer can describe the intended result and ask Claude Code to help plan or implement the technical work. It can examine documentation and text-based project data, troubleshoot an expression, write a script, build a small utility, or call operations exposed by a compatible bridge.

What it cannot do by default is open After Effects, understand an .aep file, inspect the Timeline, or move a keyframe. There is no native Connect Claude Code button in After Effects. Claude Code needs a reviewed script, a third-party extension, or an application-aware MCP server to reach the project. The quality of the result still depends on how clearly the task is scoped and how carefully the composition is reviewed.

That distinction is the key to using the combination well. Claude Code is not a replacement for After Effects or for motion-design judgment. It is a technical creative partner that can make an editable After Effects workflow faster to build, inspect, and repeat.

Why Claude Code and After Effects Are Interesting Right Now

Most generative video tools deliver a finished clip. That can be useful when the clip itself is the goal, but it is different from receiving an After Effects composition whose text, keyframes, controls, precomps, and delivery versions remain accessible.

Motion work rarely ends with the first output. Copy, timing, branding, aspect ratio, and delivery requirements change. A flattened clip may require another generation; an editable composition lets the designer revise the relevant element while preserving the rest.

Claude Code is compelling because it can operate around that editable system. Depending on the access route, it can help with several layers of work:

• Expressions: explain errors, preserve dependencies, and generate reusable property logic.

• JSX scripts: inspect projects, create compositions, add or modify layers, generate reports, and prepare Render Queue items.

• Panels and plug-ins: turn recurring studio operations into a reusable interface.

• MCP tools: expose approved After Effects actions to Claude Code and return structured results.

• Command-line rendering: help prepare and run repeatable output commands after the composition is approved.

Claude Code can turn a brief, CSV, naming convention, or production checklist into an implementation plan, explain a script before execution, and compare the returned state with the request.

The creative decision remains with the designer. Claude Code cannot determine on its own whether an entrance feels elegant, whether easing supports the message, or whether the visual hierarchy matches the brand. Even when a tool reports success, the project may still contain awkward timing, clipped text, a broken dependency, or the wrong active composition.

The useful promise is therefore not “generate a finished motion graphic with one prompt.” It is reduce the repetitive technical distance between an approved idea and an editable After Effects result.

The demand for that workflow became especially visible when Adobe announced broader Claude access to Creative Cloud tools in April 2026. Motion designer E.J. Hassenfratz replied, “Missing from this photo… After Effects.” The response had more than 88,000 views when reviewed on September 15, 2026. It was not a product demonstration, but it clearly captured the gap: designers could imagine the workflow before Adobe or Anthropic offered a documented After Effects-specific Claude connector.

Community developers and motion designers have since explored that gap with extensions, MCP servers, scripts, and custom tools. Their work offers useful evidence of what is possible, as long as a social post is treated as a demonstration rather than a reliability guarantee.

What People Are Actually Doing With Claude Code and After Effects

The most useful examples are not simply attractive final frames. They reveal how Claude reaches the application, what kind of work it performs, and whether the result remains editable.

Controlling After Effects through a third-party extension

Umair Fareed's “Using Claude Code inside After Effects to get things done” is one of the most widely shared direct examples. By September 15, 2026, the 79-second post had accumulated more than 193,000 views, 2,000 likes, and 2,100 bookmarks.

The thread supplies an important piece of context: Fareed links to AE GPT on aescripts and explains that the extension supports Claude alongside other hosted and local models. Claude is therefore reaching After Effects through a third-party application extension. The post does not reveal a native Claude Code integration.

The interaction model is nevertheless useful. A creator describes an operation, the extension translates it into actions After Effects can perform, and the creator continues working with the resulting layers and properties. The extension ultimately determines which actions and project context are available.

Showing more of the workflow, not only the reveal

Maciej Dziuba's 14-minute Claude Code and After Effects walkthrough provides more process context than a short social clip. At the time of review, the post had approximately 33,800 views and 545 bookmarks.

Its “10x more powerful” description is the creator's assessment, not a controlled benchmark. More useful is the process: what Claude Code is asked to do, how the project responds, and where human review remains necessary. A credible workflow targets named layers, preserves editable properties, exposes failures, and allows inspection before rendering.

Building an After Effects tool with coding agents

Wes McDermott showed Claude Code and Cursor being used to build an After Effects plug-in for creating materials. The post had roughly 9,500 views when reviewed.

This is not autonomous motion design. It demonstrates a potentially more durable use case: building tools that designers can review and reuse, such as a naming utility, delivery checker, lower-third builder, data importer, or project auditor. A studio can test the tool against known projects, limit what it changes, and improve it over time.

Together, these examples suggest a practical progression: start with text-based assistance, move to reviewed scripts, package proven tasks into reusable tools, and adopt a live bridge only when the additional access produces a clear benefit. A team does not need full conversational control to gain value from Claude Code.

The environment matters too. After Effects, Claude Code, the bridge, project files, dependencies, previews, and renders work most predictably on the same workstation. A Vagon Cloud Computer can provide that persistent remote environment without pretending to be the integration itself.

How to Bring Claude Code Into Your Own After Effects Workflow

The safest useful first session is deliberately small. Do not begin with a production master and a prompt such as “improve this animation.” Begin with a duplicate project, one test composition, and one outcome that can be measured.

1. Protect the source project

Duplicate the .aep file and give the working version an unmistakable name such as Product_Launch_AGENT_TEST.aep. Keep the approved original outside any directory Claude Code or a connected tool can modify.

Use separate folders for reviewed scripts, preview frames, reports, and experimental renders. If Claude Code creates layers, prefix them with something recognizable such as CC_. These boundaries make changes easier to identify and remove.

Remember that an .aep file is not ordinary source code. Claude Code should not edit its binary contents directly. After Effects should make project changes through its scripting model or through a compatible bridge operating on the open application.

2. Define the operating boundary

Place a CLAUDE.md file beside the working material and record the allowed scope. Include the project and composition names, editable layers, protected assets, output locations, naming rules, and actions that always require approval.

Project: Product_Launch_AGENT_TEST.aep
Editable comp: CC_LowerThird_Test
Editable layers: names beginning with CC_
Protected: source footage, brand assets, approved comps, existing renders
New scripts: /automation/reviewed
Preview output: /review/claude-code
Always ask before: deleting, renaming, replacing footage, saving over a file,
changing the Render Queue, or writing outside the approved folders

Project: Product_Launch_AGENT_TEST.aep
Editable comp: CC_LowerThird_Test
Editable layers: names beginning with CC_
Protected: source footage, brand assets, approved comps, existing renders
New scripts: /automation/reviewed
Preview output: /review/claude-code
Always ask before: deleting, renaming, replacing footage, saving over a file,
changing the Render Queue, or writing outside the approved folders

This file clarifies the working boundary, but it does not replace operating-system permissions or restrictions inside a third-party bridge.

3. Choose the smallest access route that fits the task

For expression troubleshooting, script generation, or report creation, Claude Code may not need live access to After Effects at all. You can supply the relevant text or project specification, review the result, and run the approved code yourself.

A JSX script is the easiest application-aware route to understand. Claude Code writes the script, you inspect it, and After Effects executes it. The workflow is less conversational than MCP, but every proposed operation remains visible before execution.

An MCP server or extension becomes useful when Claude Code needs to inspect the active project, call several approved operations, read the results back, and adapt the next step. That convenience comes with a larger trust surface: another component can now reach the open project.

4. Begin with inspection

The first connected operation should make no project changes. Ask for the After Effects version, active project and composition, dimensions, duration, frame rate, layer count, layer order, expressions, missing footage, and Render Queue state.

Inspection only. Identify the connected After Effects version, active project,
active composition, available read and preview tools, layer count, and Render
Queue state. If supported, save one preview frame to [REVIEW FOLDER]. Make no
project or file changes. Read the state again and confirm whether it changed

Inspection only. Identify the connected After Effects version, active project,
active composition, available read and preview tools, layer count, and Render
Queue state. If supported, save one preview frame to [REVIEW FOLDER]. Make no
project or file changes. Read the state again and confirm whether it changed

This smoke test confirms that Claude Code reached the intended After Effects instance and reveals what the connection can actually return. An unavailable field should be reported, not invented.

5. Request one measurable change

Once inspection works, choose a reversible property change on a named layer. Define the initial state, intended value, review frames, and stopping point.

Work only in the duplicate project [PROJECT] and composition [COMP].
Inspect all layers, but edit only [LAYER]. Do not rename, delete, replace,
or move source footage.

First report the current value and propose the smallest change. After approval,
change [PROPERTY] from [START VALUE] at frame [X] to [END VALUE] at frame [Y].
Preserve existing parenting and controls. Read the changed values back, render
review frames at [TIMES] to [NEW OUTPUT FOLDER], then stop for review

Work only in the duplicate project [PROJECT] and composition [COMP].
Inspect all layers, but edit only [LAYER]. Do not rename, delete, replace,
or move source footage.

First report the current value and propose the smallest change. After approval,
change [PROPERTY] from [START VALUE] at frame [X] to [END VALUE] at frame [Y].
Preserve existing parenting and controls. Read the changed values back, render
review frames at [TIMES] to [NEW OUTPUT FOLDER], then stop for review

The prompt creates a controlled loop: inspect, propose, approve, modify, read back, preview, and stop. Success is measurable, and the designer retains the decision point.

6. Expand only after the loop is reliable

Once the first task behaves predictably, the same pattern can support larger jobs:

• repair a known expression issue across selected layers

• create a standard composition with exact dimensions, duration, and frame rate

• generate an approved group of text, shape, null, or control layers

• apply one property change across explicitly named layers

• create a data-driven version from a master composition and controlled CSV

• duplicate an approved master into 9:16 or 1:1 delivery compositions

• prepare preview frames or Render Queue items for human review

For batch work, generate one test version first. Check copy overflow, missing assets, duration, naming, and output location before approving the remaining variants. For alternate aspect ratios, duplicate the master into new compositions rather than resizing the approved source in place.

For expressions, preserve the original text and identify dependencies such as layer names, controls, parents, markers, or referenced properties. Removing the visible error is not enough if the new expression silently breaks the rig.

Separate creative intent from technical execution. You decide how a title should enter, accelerate, and settle. Claude Code can translate that approved decision into keyframes or expressions. Asking the same system to invent, implement, and approve the motion in a single step removes the checkpoints that make the workflow dependable.

Review the Project, Not Only the Tool Response

A successful command means that the tool returned success. It does not prove that the correct composition changed or that the animation looks right.

After every material operation, read the affected layers, properties, keyframes, expressions, and output paths back from After Effects. Compare those values with the approved request. Then inspect the actual composition at the beginning, during active motion, and after the animation settles.

Still frames reveal cropping, overlap, alignment, and unexpected visibility. To judge rhythm, play the sequence at its intended frame rate and review easing, pauses, and continuity.

Before final delivery, verify:

• composition dimensions and frame rate

• work-area and output duration

• required codec, alpha, and audio settings

• fonts, footage, plug-ins, and expression dependencies

• output name and destination

• playback of the rendered file, not only the composition preview

Adobe's official aerender can automate output from an already prepared project. Claude Code can help construct or troubleshoot the command, but aerender does not design or visually approve the composition.

No Native Claude Code Connector? Use the Route That Matches the Job

As of September 2026, Adobe and Anthropic do not document an official After Effects-specific MCP server that lets Claude Code control an open project. That does not make the workflow impossible. It means the connection method must be selected and evaluated separately.

For most teams, reviewed scripts are the sensible entry point. Begin with report-only scripts. Move to reversible edits wrapped in an undo group. Keep the approved project untouched until the workflow has been tested on representative files.

For a more fluid conversation with an open project, one community option identified during this research is JUNKDOGE-JOE/after-effects-mcp. Its documented setup combines a CEP panel with a local MCP endpoint and describes project readback, preview, expression-validation, checkpoint, and revert operations. Its Claude Code connection command is:

claude mcp add --transport

claude mcp add --transport

Treat any community bridge as software with access to valuable project material. Review its source and release information, verify the package you install, pin the tested version, inspect the exposed tool list, and begin with a duplicate project. A maintainer's test results are not an Adobe, Anthropic, or Vagon guarantee.

If Claude Code is not a requirement, other products may offer a more packaged experience. Adobe's AI Assistant for After Effects Beta is the official Adobe route, while Higgsfield markets a commercial prompt-to-motion workflow that keeps elements editable. These are alternatives to evaluate, not evidence of a native Claude Code connection, and their availability, supported operations, pricing, and service requirements may change.

Use reviewed scripts when transparency matters most, add MCP when live readback justifies the additional trust, and choose a packaged alternative when convenience matters more than controlling the toolchain.

How Vagon Makes the Workflow Easier to Run

The connection between Claude Code and After Effects is only one part of the workflow. The complete setup may also depend on a licensed After Effects installation, a compatible extension or MCP panel, Node or scripting dependencies, fonts, plug-ins, source footage, disk cache, preview files, and a render destination.

When those parts are divided between a laptop, another workstation, and a render machine, troubleshooting becomes harder. The bridge may point to the wrong application instance. A font or plug-in may be missing. Local cache and project files may not match. Long operations remain tied to the computer sitting in front of you.

Vagon Cloud Computer gives the workflow one persistent remote Windows workstation. You install and manage After Effects, Claude Code, the chosen bridge, project dependencies, and licensed creative tools in that environment. Project files, cache, previews, and renders can remain beside the applications that use them.

That is especially useful for a local MCP route. A local endpoint generally expects the agent client and the After Effects panel to be available on the same computer. Running both inside one Vagon Cloud Computer avoids turning the bridge into a connection between separate machines. You remotely control the same workstation on which the project is open and the tools are running.

A practical setup looks like this:

1. Create a Vagon Cloud Computer appropriate for the project.

2. Install your licensed copy of After Effects, Claude Code, plug-ins, fonts, and approved bridge.

3. Place the duplicate project and required source files in the working environment.

4. Define protected folders and separate locations for scripts, reviews, and renders.

5. Run the inspection and single-change workflow.

6. Review the actual Timeline, preview, and rendered output from the same workstation.

7. Increase resources for a heavier composition or render, then return to a suitable configuration afterward.

This does not make an untrusted integration safe or turn a vague prompt into good creative direction. It removes a different source of friction: keeping the applications, dependencies, project state, and compute resources together while you access them remotely.

Choose the workstation around the project. Adobe's current After Effects system requirements list 16 GB of RAM as the Windows minimum, recommend 32 GB or more for 4K work, recommend 8 GB of GPU memory, and call for fast SSD storage for the application and cache. High-resolution footage, layer-heavy compositions, Multi-Frame Rendering, and GPU-intensive effects may require more.

Vagon provides the remote computer and working environment. It does not include Adobe Creative Cloud, an After Effects license, Claude Code, Anthropic usage, an MCP server, third-party extensions, plug-ins, fonts, codecs, or external AI credits. You retain control of the installed software, project access, automation boundary, and final creative approval.

Frequently Asked Questions

Does Claude Code integrate natively with After Effects?

No. Claude Code currently needs a reviewed script, compatible third-party extension, or community After Effects MCP bridge to interact with an open project. Adobe's AI Assistant is a separate official beta feature.

What is the easiest way to start?

Ask Claude Code to explain an expression or produce a report-only JSX script for a duplicate test project. Review the script before running it. Move to a live MCP connection only when you need interactive inspection, readback, and repeated operations.

Can Claude Code edit an .aep file directly?

It should not edit the binary file as ordinary text. Let After Effects perform changes through its scripting model or a compatible bridge, and keep the approved source outside the editable workspace.

Does After Effects need to remain open?

Usually yes for a panel, interactive MCP workflow, or script operating on the active application. aerender can launch a rendering instance for an already prepared project without the standard interactive interface.

Can Claude Code create a complete motion graphic automatically?

A capable bridge may expose enough operations to build compositions, layers, keyframes, and expressions. That does not guarantee good timing, hierarchy, continuity, or brand fit. Claude Code can implement a defined motion plan; a motion designer should review and approve it.

Is a community After Effects MCP server safe?

Do not assume that it is. Review the source and release, inspect its tools, limit accessible folders and operations, pin the tested version, and use a duplicate project. Claude Code's permissions do not automatically constrain every action exposed by third-party software.

Can I use Codex instead of Claude Code?

Potentially. Some community bridges document more than one MCP-compatible agent route. Compare the tools with the same duplicate project, task, access boundary, and review criteria. The quality of the bridge and the clarity of the workflow matter as much as the selected coding agent.

Does Vagon include After Effects or Claude Code?

No. Vagon provides the remote Windows computer. You install and manage your own licensed applications, accounts, plug-ins, bridge software, and AI usage.

Build the Workflow on One Cloud Workstation

Claude Code can make After Effects automation easier to plan, write, and operate, but the dependable version of the workflow is still controlled: protect the source, expose only the required operations, approve narrow changes, inspect the project, and review the rendered result.

When you want After Effects, Claude Code, the selected connection, project files, dependencies, previews, and rendering resources to remain together, run the setup on a Vagon Cloud Computer. You can access the same creative workstation remotely, choose resources for the project, and keep control of the applications and final output.

Create your Vagon Cloud Computer