Skip to main content

Cavi After Effects MCP

License: MIT

MCP server for headless Adobe After Effects on macOS. It starts After Effects without its UI, runs ExtendScript with arguments and returns the script's JSON report, and renders saved projects with aerender, checking every output movie's frame count. 45 tools cover comps, layers, shapes, masks, text, keyframes, effects, 3D, transitions, templates, Essential Graphics and delivery.

Try one title

Put a lower third over a shot: the footage pushes in, a name bar slides in, and two lines of text rise in. The frame below was rendered by headless After Effects from that comp at 2.5 seconds.

A server aisle in blue and magenta light with a lower third reading "Aisle B" and "Racks online at 02:00"

The calls are in run.py and the shot is in examples/first-result/media. The comp is 1280 by 720, 3.5 seconds at 24 fps, with five layers. Reproduce the title with one command; the calls are explained in Your first comp.

Install

Requires macOS, After Effects 2026 or later, FFmpeg (ffmpeg and ffprobe on PATH) and Python 3.15 or later with uv. You supply your own licensed Adobe applications.

uv tool install cavi-after-effects-mcp

Upgrade with uv tool upgrade cavi-after-effects-mcp. To work from a checkout instead, see CONTRIBUTING.md.

Quickstart

Register the stdio server with your MCP client:

{"mcpServers": {"after-effects": {"command": "cavi-after-effects-mcp"}}}

If your client does not inherit your shell's PATH, use the absolute path that command -v cavi-after-effects-mcp prints. Then ask your agent to:

  1. Call ae_status. It reports the After Effects it found and its version, or the reason it found none.
  2. Call ae_launch. It starts one headless instance and waits until it answers.
  3. Create a comp with ae_comp_new, passing an absolute path for project. The project is created and saved.
  4. Add layers, animation and effects, then look with ae_frame or write a movie with ae_export_frames.
  5. Render a queue with ae_render_queue_set and ae_render_queue.

Every tool that edits a project takes the project's absolute path and saves it. After Effects runs with no UI, so nothing is open on screen; a call that fails leaves the project as it was.

What it does

Area Tools
Session ae_status, ae_launch, ae_script
Project ae_inspect, ae_project, ae_project_settings, ae_import, ae_footage_replace, ae_collect, ae_organize, ae_fonts, ae_effects, ae_presets
Comps and layers ae_comp_new, ae_comp_set, ae_layer_add, ae_layer_set, ae_layer_edit, ae_text_set, ae_shape_add, ae_mask, ae_camera, ae_material, ae_markers
Animation ae_keyframes, ae_expressions, ae_effect_add, ae_preset_apply, ae_text_animator, ae_controls, ae_audio, ae_time, ae_properties
Transitions ae_transitions, ae_transition, ae_sequence_layers
Delivery ae_render_queue_set, ae_render_queue, after_effects_render, ae_frame, ae_export_frames, ae_template_render, ae_essential_graphics, ae_essential_override, ae_mogrt_export

Each tool carries its access in its annotations: read-only, write, or destructive. The tool reference lists every tool's parameters and result fields, and a test fails when it and the server's registry differ. Three calls always refuse, because headless After Effects aborts on the underlying operation, and name the supported alternative instead: ae_preset_apply, ae_mogrt_export and convert_to_keyframes of ae_audio. See Measured behavior.

Script contract

ae_script(script, project, args) defines the global CAVI_ARGS (the args object), then runs the script. The script sets its report with caviReport(object), which stores the JSON in the saved project. A script that never calls caviReport reports {"value": <its last value>}. A report with a non-empty errors list fails the call with those errors. Before running, the runner compiles the script inside try, so a syntax error comes back as a message with its line. A script that fails leaves no partial edits.

var comp = app.project.items.addComp("Title", 1920, 1080, 1, 2, 30);
caviReport({comp: comp.name, frames: 60, errors: []});

ae_script refuses scripts that call applyPreset, executeCommand or findMenuCommandId, which abort headless After Effects.

Rendering

ae_render_queue runs aerender on a saved project's queue and checks each declared output's frame count with ffprobe. A movie that is missing, older than the render or short of frames is rendered again on its own. If that still fails, the movie is encoded from saveFrameToPng frames with ffmpeg and the output reports which path produced it. ae_export_frames writes a frame range without aerender at all. See Rendering and the fallback.

after_effects_render renders one declared composition range of a workspace project to a new PNG directory and writes a receipt. The recipe:

{
  "version": 1, "name": "Title", "project": "titles/title.aep",
  "dependencies": ["titles/aisle.mp4"], "composition": "Title", "start": 0, "end": 83,
  "fps": {"numerator": 24, "denominator": 1}, "loop": false, "alpha": "straight",
  "output_module": "RGBA PNG"
}

Start and end are inclusive frames. The output module must be an existing straight-alpha RGBA PNG template, and exactly one aerender must be installed.

Documentation

The documentation source is in docs/after-effects-mcp/source:

Configuration

Variable Meaning
CAVI_AFTER_EFFECTS_APP Absolute path of the After Effects .app to use. A wrong path is reported, never replaced by another install. Default: the newest Adobe After Effects 20* in /Applications.
CAVI_ADOBE_WORKSPACE Workspace folder for after_effects_render; its recipe and output must stay inside it.
CAVI_ADOBE_DATA Private state directory of the shared Adobe package.

Security

The server speaks MCP over stdio, sends no telemetry and uses no accounts. It runs ExtendScript with the permissions of the user who started it. Only after_effects_render is confined to a workspace folder; the other tools take any absolute path and can write there, and ae_script runs arbitrary ExtendScript. Run it only for clients you trust. See the security model and SECURITY.md.

Tests

uv sync --locked --dev
uv run --locked ruff check .
uv run --locked pytest -q
uv run --locked python evals/run.py

The unit tests and the evaluations drive the registered tools through the server and never start After Effects; tests that need FFmpeg skip without it. A test fails when the tool reference and the registry differ, when the documentation build is not reproducible, or when a link in the README or the pages is broken. The evaluations play each case in evals/cases.json against the server over stdio. See Testing and evaluation.

Contributing

See CONTRIBUTING.md for setup, checks and the rules for changes. This project follows the Code of Conduct.

License

MIT. See LICENSE.

Trademarks

Adobe and After Effects are either registered trademarks or trademarks of Adobe in the United States and/or other countries. This project is not affiliated with, endorsed by, or sponsored by Adobe. It does not include or redistribute Adobe software, templates or presets; you supply your own licensed applications.

Metadata

Release files for cavi-after-effects-mcp 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for cavi-after-effects-mcp 0.1.0
File Size Uploaded Provenance Actions
cavi_after_effects_mcp-0.1.0.tar.gz 90.1 kB With provenance Details for cavi_after_effects_mcp-0.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for cavi-after-effects-mcp 0.1.0
File Interpreter ABI Platform Actions
cavi_after_effects_mcp-0.1.0-py3-none-any.whl Python 3 none any Details for cavi_after_effects_mcp-0.1.0-py3-none-any.whl

Total release size: 175.2 kB

Release files / cavi_after_effects_mcp-0.1.0.tar.gz

Download URL cavi_after_effects_mcp-0.1.0.tar.gz
Size 90.1 kB
Tags Source
SHA-256 checksum
How to use checksums
e9880008a1ac3c3a0e819f75a7b365d54cb235824b3af798abe794277b8a85f2
BLAKE2b-256 checksum
How to use checksums
3ce50ad6d6249d18c0a7c6aae02a0aaf5a291611a494bde8d2feab3c383e7a7c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 14, 2026.

Transparency log

Release files / cavi_after_effects_mcp-0.1.0-py3-none-any.whl

Download URL cavi_after_effects_mcp-0.1.0-py3-none-any.whl
Size 85.1 kB
Tags Python 3
SHA-256 checksum
How to use checksums
1b58ca9380c5526f7d8fe9ea07aad7b313ce7e3bd072f9274530259d396e0129
BLAKE2b-256 checksum
How to use checksums
5d615ff868206bae36b383ef280148e51d34d28b229f3af76dbe332313f07d6b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 14, 2026.

Transparency log

Security for cavi-after-effects-mcp 0.1.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for cavi_after_effects_mcp-0.1.0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 14, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files Full provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page