FFmpeg MCP Server
Enables video processing tasks such as resizing, audio extraction, and metadata inspection using FFmpeg, with base64 input/output for stateless and portable operation.
README
FFmpeg MCP Server
MCP server providing FFmpeg video processing via base64 I/O. Stateless, portable, zero external dependencies.
Tools
get_ffmpeg_version
Returns FFmpeg version information.
resize_video
Resize video to standard resolutions (360p, 480p, 720p, 1080p). Outputs H.264/AAC.
Parameters:
videoBase64(string): Base64 encoded videoresolutions(array): Target resolutionsoutputFormat(string, optional): mp4 or webm (default: mp4)
Returns: Base64 encoded video for each resolution
extract_audio
Extract audio track from video.
Parameters:
videoBase64(string): Base64 encoded videoformat(string): mp3, aac, wav, or ogg
Returns: Base64 encoded audio
get_video_info
Inspect video metadata: format, duration, bitrate, resolution, codecs, streams.
Parameters:
videoBase64(string): Base64 encoded video
Returns: Formatted metadata
Design
Base64 I/O: All inputs and outputs are base64 encoded. No file paths, S3 URLs, or database storage.
Why:
- Stateless: No persistent storage, cleanup, or orphaned files
- Portable: Works in any Node.js environment
- Simple: Single request/response cycle
Constraints:
- Payload limits: ~4-6 MB depending on runtime
- Best for: Thumbnails, previews, short clips, audio extraction
- Not for: Feature films, 4K processing
Processing flow:
- Decode base64 input → temp file
- Execute FFmpeg
- Read output → buffer
- Delete temp files
- Encode buffer → base64
Installation
FFmpeg binaries install automatically via ffmpeg-static and @ffprobe-installer/ffprobe.
npm install @windsornguyen/ffmpeg-mcp
Configuration
Claude Desktop
claude_desktop_config.json:
{
"mcpServers": {
"ffmpeg": {
"command": "npx",
"args": ["-y", "@windsornguyen/ffmpeg-mcp"]
}
}
}
VS Code
.vscode/mcp.json:
{
"servers": {
"ffmpeg": {
"command": "npx",
"args": ["-y", "@windsornguyen/ffmpeg-mcp"]
}
}
}
Development
# Install dependencies
bun install
# Build
bun run build
# Test
bun test
# Run STDIO transport
bun run dev:stdio
# Run HTTP transport
bun run dev:shttp
Environment Variables
PORT: HTTP server port (default: 3002)NODE_ENV: development | production
Testing
# Run all tests
bun test
# Run with output artifacts (kept in tests/output/)
bun test --keep-artifacts
# Single test file
bun test tests/ffmpeg.test.ts
Tests verify base64 round-trip encoding, resolution accuracy, and codec compliance. Output artifacts saved to tests/output/ when --keep-artifacts is set.
Architecture
- Client (
src/client.ts): FFmpeg execution with base64 I/O - Tools (
src/tools/ffmpeg.ts): MCP tool definitions - Server (
src/server.ts): Request routing - Transport (
src/transport/): STDIO and HTTP support
License
MIT License. See LICENSE file for details.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。