Vlexor MCP Server
Enables MCP agents to drive AI video editing on local footage, game-asset generation, and motion-graphics production via Vlexor's API, with all video rendering done entirely locally.
README
@velxor/mcp
An MCP (Model Context Protocol) server that lets Claude Code, Cursor, and other MCP-capable agents drive all three Vlexor pillars — AI video editing, game-asset generation, and motion-graphics production — directly from a coding session.
Video editing runs on your own footage: your raw video never leaves your machine — only the text of your instructions (and an optional transcript) is sent to Vlexor to generate the edit plan, which is compiled to a real FFmpeg command and rendered entirely locally. Asset generation and motion graphics render server-side (there's no local footage to protect) and return a signed download URL — optionally saved straight to a local path.
Setup
- Generate a personal API key at Vlexor → Settings → MCP & API.
- Install FFmpeg if you don't already have it (
brew install ffmpegon macOS, or pointVLEXOR_FFMPEG_DIRat a folder that hasffmpeg/ffprobe). - Add this to your MCP client's config:
{
"mcpServers": {
"vlexor": {
"command": "npx",
"args": ["-y", "@velxor/mcp"],
"env": { "VLEXOR_API_KEY": "vlx_…" }
}
}
}
Tools
Pillar 1 — video editing (renders locally; footage never uploaded):
probe_media— read a local video's duration, resolution, fps, codec.generate_edit_plan— natural-language instruction → structured edit plan (calls the real Vlexor Command Engine; counts against your plan's usage).compile_edit_plan— edit plan → a real FFmpeg command, computed entirely locally.render— run the compiled command locally to produce the final video.
A typical agent flow: probe_media → generate_edit_plan → (review the
steps) → compile_edit_plan → (review the FFmpeg command) → render.
Pillar 2 — asset generation (renders server-side):
generate_game_asset— prompt → a texture, 2D character, sprite sheet, 3D model (.glb), sound effect, music track, parallax background, or character turnaround. Returns a signed download URL; passsavePathto also write the file locally.
Pillar 3 — motion graphics (renders server-side):
generate_motion_graphic— a short creative brief → a rendered kinetic-title.mp4(announcement/promo card). Returns a signed download URL; passsavePathto also write the file locally.
Environment variables
| Variable | Required | Default | Purpose |
|---|---|---|---|
VLEXOR_API_KEY |
yes | — | Personal API key from Settings → MCP & API |
VLEXOR_API_BASE |
no | https://vlexor.app |
Override for self-hosted/dev instances |
VLEXOR_FFMPEG_DIR |
no | (system PATH) |
Folder containing ffmpeg/ffprobe, if not on PATH |
About this repository
This package is the thin integration layer that connects an MCP client to
Vlexor's real, hosted API — every tool call here is a plain, inspectable
HTTP request (you can read exactly what leaves your machine in
src/api-client.ts). It's open so agent developers
piping their API key through it can verify that for themselves.
The Vlexor platform itself — the editing engine, the AI prompt design, the
asset/motion-graphics generation pipelines — is closed-source. A couple of
this package's imports (@velxor/engine's FFmpeg-plan compiler,
@velxor/shared's type definitions) come from that private codebase and
aren't published separately, so this repo won't pnpm install/build
standalone — it's meant to be read, and run via the pre-built npm package
(npx -y @velxor/mcp), not compiled from this checkout.
Issues and PRs on the MCP integration layer itself (tool schemas, error handling, the local FFmpeg orchestration) are welcome.
License
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。