Curated, agent-first directory of MCP servers for AI filmmaking.
概览
Every tool your agent needs to make a film — video models, voices, scores, edit bays, finishing suites, and the pipes to ship it. Curated, verified, agent-first, and re-checked daily. 🎬 mcp.film · 🤖 /llms.txt · /api/registry.json · claude mcp add mcp-film -- npx -y mcp-film A directory of verified Model Context Protocol servers across the full film pipeline — from screenplay breakdown to YouTube upload — each entry annotated with exact install commands, auth requirements, pricing, sample tools, and the caveats that actually bite ("requires Resolve Studio", "Suno has no official API", "uploads cost 1,600 quota units"). Feedback typed on a server page is captured as a PostHog event for a maintainer to read; nothing is folded into the rankings automatically. Rankings are editorial: official status, maintenance recency, hosted availability, and the caveats each entry carries. Every change is a commit — the entire editorial history is auditable.
README
以下为原文 README 节选，完整内容请查看原文。
mcp.film
The MCP directory for AI filmmaking. Every tool your agent needs to make a film — video models, voices, scores, edit bays, finishing suites, and the pipes to ship it. Curated, verified, agent-first, and re-checked daily.
🎬 Live site: mcp.film ·
🤖 For agents: /llms.txt · /api/registry.json ·
claude mcp add mcp-film -- npx -y mcp-film
What this is
A directory of verified Model Context Protocol servers across the full film pipeline — from screenplay breakdown to YouTube upload — each entry annotated with exact install commands, auth requirements, pricing, sample tools, and the caveats that actually bite (“requires Resolve Studio”, “Suno has no official API”, “uploads cost 1,600 quota units”).
It is built agents-first:
| Surface | URL |
|---|---|
| llms.txt index (spec) | /llms.txt · /llms-full.txt |
| Full structured registry | /api/registry.json (+ .min.json) |
| MCP Registry-compatible API | /v0.1/servers · /api/mcp-registry.json |
| Catalog freshness / ops pulse | /api/pulse.json · /pulse.md |
| Production playbooks | /api/playbooks.json · /playbooks.md |
| Agent recommendations | /api/recommendations.json · /recommendations.md |
| Capability index | /api/capabilities.json · /capabilities/{tag}.md |
| Client setup profiles | /clients/ · /clients.md · /api/client-profiles.json |
| One server, JSON / markdown | /api/mcps/{slug}.json · /mcps/{slug}.md |
| Pipeline guide | /stack/ · /stack.md |
| New servers feed | /feed.xml |
| The directory as an MCP server | npx -y mcp-film (packages/mcp-server/) |
JSON-LD (SoftwareApplication, ItemList) |
embedded in every page, server-rendered |
How it runs itself
┌────────────────────────────────────────────────┐
│ data/ (the product) │
└───────▲────────────────────────────────────────┘
daily: re-verify │
links, hunt new │
servers │
┌─────────────┴──┐
│ curator.yml │
│ (Claude agent) │
└─────────────┬──┘
│
▼
PRs labeled `auto-data` ──► auto-merge.yml
(merges ONLY if every changed file is in data/
and validation passes) ──► deploy.yml ──► Cloudflare Pages
└──► GitHub Pages fallback
Feedback typed on a server page is captured as a PostHog event for a maintainer to read; nothing is folded into the rankings automatically. Rankings are editorial: official status, maintenance recency, hosted availability, and the caveats each entry carries. Every change is a commit — the entire editorial history is auditable.
Agent-readable traffic is measured at the edge when deployed on Cloudflare
安装
This server does not publish a one-line install command.
Open the repository installation guide 新手？先看 mcpskill.org 教程 →