Media-Editor-MCP
Provides MCP servers for AI-driven video editing. Enables offline editing of CapCut drafts and remote control of Adobe Premiere Pro via UXP plugin, with shared media analysis for beat detection and transcription.
README
Media-Editor-MCP
A monorepo of MCP servers for AI-driven video editing, plus the shared media-analysis server they both rely on.
Packages
| Package | Lang | What it is |
|---|---|---|
packages/capcut-mcp |
Python / uv | Edits CapCut drafts on disk (offline): clips, audio, text/captions, effects, filters, transitions. |
packages/premiere-server |
TypeScript | MCP server that drives Adobe Premiere Pro via a UXP plugin bridge. |
packages/premiere-uxp-plugin |
TypeScript | The UXP plugin loaded inside Premiere; the server talks to it over WebSocket. |
packages/premiere-protocol |
TypeScript | Shared message/types contract between the Premiere server and plugin. |
packages/media-analysis |
Python / uv | Editor-agnostic analysis: beat/energy detection, Whisper transcription, SRT, best-moments, stock fetch. Used by both editors. |
Layout & tooling
Polyglot monorepo — unified by git, Claude skills, and one MCP config, not a single build:
- TypeScript (
premiere-*) uses npm workspaces (rootpackage.json). Build:npm install && npm run build. - Python packages each keep their own uv venv — they are not a shared environment.
media-analysispulls Whisper/torch (heavy);capcut-mcpis light and must not inherit those deps. capcut-mcpdoes not importmedia-analysis; it calls it as a separate MCP server. The only coupling is registration in.mcp.json.
.
├─ packages/
│ ├─ capcut-mcp/ (python/uv) src/ccmcp
│ ├─ media-analysis/ (python/uv) src/media_analysis
│ ├─ premiere-server/ (ts) @ppmcp/server
│ ├─ premiere-uxp-plugin/ (ts) @ppmcp/uxp-plugin
│ └─ premiere-protocol/ (ts) @ppmcp/protocol
├─ .claude/skills/ cc-* (CapCut) and pp-* (Premiere) workflow skills
├─ .mcp.json.example sanitized server registrations (copy to .mcp.json, add keys)
├─ package.json npm-workspaces root (TS only)
└─ scripts/
History & contribution model
capcut-mcp and the Premiere stack were each imported once via git subtree add
(then restructured into packages/). This monorepo is the single source of truth —
the full history of both originals is preserved here in main.
- Work on
main. No subtrees, no submodules, no per-package branches. - Do not run
git subtree pull/pushagainst the old standalone repos. The Premiere code was moved out of its original_pp/prefix, so the prefixes no longer align and a sync would corrupt the tree. Treat the original capcut/premiere repos as archived. - New work in any package is a normal commit here; there is nothing to sync upstream.
Setup
# TypeScript (Premiere) deps + build
npm install
npm run build
# Python servers — independent venvs (run once each)
uv sync --directory packages/capcut-mcp
uv sync --directory packages/media-analysis
# MCP registration: copy the template, add your Pexels/Pixabay keys
cp .mcp.json.example .mcp.json
Open Claude Code at this repo root so the relative paths in .mcp.json resolve. .mcp.json is
gitignored (it holds API keys); share changes via .mcp.json.example.
Per-editor profiles (lighter context)
All three servers in .mcp.json load their tool schemas into every session (~12–18k tokens),
but you never edit in CapCut and Premiere at once. Two slimmer profiles each load just one
editor plus the shared media-analysis:
# CapCut session — skips Premiere's ~6–8k tokens of tool schemas
claude --strict-mcp-config --mcp-config .mcp.capcut.json
# Premiere session — skips CapCut's tool schemas
claude --strict-mcp-config --mcp-config .mcp.premiere.json
--strict-mcp-config makes Claude Code use only the named file, ignoring the default
.mcp.json and user/global servers. Both profiles are gitignored (they hold API keys); copy
the .mcp.capcut.json.example / .mcp.premiere.json.example templates and add your keys. Use
plain .mcp.json (all servers) when you want everything loaded.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.