v1.2.4 — 81 tools tested & working

Video editing
for AI agents.

An open-source MCP server that gives Claude Code, Cursor, and any AI agent 81 video editing tools via FFmpeg and Remotion. Local, fast, and completely free.

81
Tools
858
Tests
279
Commits
0$
Forever
View on GitHub pip install mcp-video
mcp-video
$ mcp-video trim input.mp4 -s 0:30 -d 15 -o clip.mp4
✓ clip.mp4 — 15.0s @ 1920x1080
$ mcp-video filter clip.mp4 -t grayscale -o bw.mp4
✓ bw.mp4 — grayscale applied
$ mcp-video add-text bw.mp4 "Hello World" -p center --size 48 -o final.mp4
✓ final.mp4 — text rendered
# Or use the Python client
$ python3 -c 'from mcp_video import Client; c=Client(); c.trim("in.mp4", start=5, duration=10)'
✓ Trimmed to 10s — output: trimmed.mp4
Features
Everything you need to edit video with AI.

81 tools across 9 categories. Every tool returns structured JSON with output paths, duration, and resolution. On failure, you get clear error messages with auto-fix suggestions.

10
Core Editing
8
Filters
5
Effects
7
Audio
10
AI Features
7
Remotion
✂️

Core Editing

Trim, merge, crop, rotate, resize, convert, speed change, fade, watermark. The basics, done right.

10 tools
🎨

Filters & Effects

Blur, sharpen, grayscale, sepia, vignette, chromatic aberration, scanlines, noise, glow.

13 tools
🎬

Transitions

Glitch, pixelate, and morph transitions for seamless clip connections.

3 tools
🎵

Audio Suite

Normalize, reverb, compress, pitch shift, noise reduction. Extract waveforms and generate sound effects.

7 tools
🤖

AI-Powered

Whisper transcription, scene detection, stem separation, AI upscaling, color grading, spatial audio.

10 tools
📐

Layout & MoGraph

Grid layouts, picture-in-picture, split-screen, animated text, count animations, progress bars.

8 tools
Why mcp-video
Not just another FFmpeg wrapper.

Raw FFmpeg is powerful but painful. Cloud APIs are easy but expensive. mcp-video gives you the power of FFmpeg with the ergonomics of a modern API.

mcp-video Raw FFmpeg Cloud APIs
Cost ✓ Free, local ✓ Free ✗ $0.28–$2.50 per 5s
Privacy ✓ 100% local ✓ 100% local ✗ Uploads to cloud
Error handling ✓ Structured + auto-fix ✗ Cryptic stderr ✗ HTTP errors
Validation ✓ Before execution ✗ Runtime only ✓ API schema
Text overlay add_text("Hi") -vf "drawtext=text='Hi':fontfile=..." Varies
Agent discovery ✓ 81 self-documenting tools ✗ Memorize flags ✗ Read docs
Testing ✓ 858 tests ✗ None Provider's
Tools
81 tools. Every operation you need.

All tested end-to-end on real MP4s. Every tool returns structured JSON with output paths, duration, and resolution.

✓
video_info
Duration, resolution, codec, fps, bitrate
✓
video_trim
Trim by start/duration or end time
✓
video_merge
Concatenate clips with transitions
✓
video_add_text
Overlay text with positioning
✓
video_add_audio
Mix, replace, or add audio tracks
✓
video_resize
Change resolution or aspect ratio
✓
video_convert
mp4, webm, gif, mov with presets
✓
video_speed
Speed up or slow down
✓
video_filter
blur, grayscale, sepia, sharpen, etc.
✓
video_crop / rotate
Crop regions, rotate, flip
✓
video_fade / watermark
Fade in/out, image watermarks
✓
video_subtitles
Burn SRT/VTT, generate from text
✓
video_thumbnail / preview
Extract frames, low-res preview
✓
video_storyboard
Key frame grid for review
✓
video_stabilize
Motion vector stabilization
✓
video_overlay / split_screen
PIP, side-by-side layouts
✓
video_detect_scenes
Auto scene change detection
✓
video_layout_grid / pip
Multi-video grid layouts
✓
video_text_animated
Animated text overlays
✓
video_mograph_count / progress
Counter animations, progress bars
✓
audio_preset / sequence
Synthesized sound effects
✓
audio_compose / effects
Layer tracks, apply reverb, etc.
✓
video_audio_spatial
3D audio positioning
✓
video_ai_transcribe
Whisper speech-to-text
✓
video_ai_stem_separation
Isolate vocals, drums, bass
✓
video_ai_upscale
ML super-resolution (2x/4x)
✓
video_ai_scene_detect
ML scene boundary detection
✓
video_ai_color_grade
Auto color correction
✓
video_ai_remove_silence
Auto-remove dead air
✓
remotion_render / still
React-based video rendering
✓
search_tools
Search tools by keyword
✓
...and 49 more
See GitHub for full list
Three Interfaces
However you want to use it.

MCP server for AI agents. Python library for scripts. CLI for the terminal. All three share the same 81 tools.

MCP Server

Add to Claude Code or Cursor. Then just talk to your agent.

JSON 
{
  "mcpServers": {
    "mcp-video": {
      "command": "uvx",
      "args": ["mcp-video"]
    }
  }
}

Python Client

Clean API for automation and pipelines.

Python 
from mcp_video import Client

editor = Client()
clip = editor.trim("v.mp4", start="0:30", duration="15")
final = editor.resize(clip.output_path, aspect_ratio="9:16")
result = editor.export(final.output_path)
print(result.resolution)  # 1080x1920

CLI Tool

Human-friendly terminal output. Add --format json for scripting.

Bash 
# Info with rich table output
$ mcp-video info video.mp4

# Trim with progress
$ mcp-video trim video.mp4 -s 0:30 -d 15

# Apply TikTok template
$ mcp-video template tiktok video.mp4 --caption "Hello!"

# JSON output for scripts
$ mcp-video --format json info video.mp4
Install
Get started in 30 seconds.

Requires Python 3.11+ and FFmpeg installed. AI features need optional dependencies.

Base Install
Terminal 
pip install mcp-video
With AI Features
Terminal 
pip install mcp-video[ai]
Verify Installation
Terminal 
mcp-video --version
mcp-video doctor  # Check FFmpeg and optional deps
Quality
858 tests. Real media where it matters.

Core video paths use actual FFmpeg operations. Every tool was tested end-to-end on real MP4s.

858
Tests Collected
817
Passing
81
Tools Verified
7
Bugs Found & Fixed

🎯 Real FFmpeg Operations

Core video paths invoke actual FFmpeg subprocesses — no mocks. If it passes, it works on your machine.

🔧 Auto-Fix Error Handling

FFmpeg errors are parsed into structured responses with actionable suggestions. No more cryptic stderr.

🛡️ Input Validation

Filter types, color presets, and parameters are validated before hitting FFmpeg. Fail fast, fail clear.

📊 Structured Output

Every tool returns JSON with output_path, duration, resolution, bitrate, and codec. Agents always know what happened.