Reel Estate MCP
Turn real-estate listing photos into cinematic property videos from your AI assistant. reel-estate-mcp is a Model Context Protocol (MCP) server that connects Claude, Cursor, and other AI assistants to your Reel Estate account — so you can generate motion clips, AI-edit photos (virtual staging, twilight, seasonal), add voiceover, and render finished listing videos, all in plain language.
README
reel-estate-mcp
Turn real-estate listing photos into cinematic property videos from your AI
assistant. reel-estate-mcp is a Model Context Protocol
(MCP) server that connects Claude, Cursor, and other AI assistants to your
Reel Estate account — so you can generate motion
clips, AI-edit photos (virtual staging, twilight, seasonal), add voiceover, and
render finished listing videos, all in plain language.
Reel Estate is the AI real-estate video platform that turns property photos into scroll-stopping listing videos. Create a free account at tryreelestate.com.
You sign in once through your browser (OAuth + PKCE — no API keys, no secrets),
and every call is proxied through the backend's /mcp endpoint, which stays the
single authority for auth, plans, and permissions. This bridge's one superpower on
top of that: it can read local image files and stream them straight into a
project — something a purely remote server can't do.
- 🎬 Generate real-estate videos — animate photos into clips, then render the movie
- 🖼️ AI photo editing — virtual staging, twilight, upscale, seasonal, replace/remove/add
- 🎙️ Voiceover & timeline — narration, music, overlays, reordering
- 📤 Local uploads — push photos from disk into a project
- 🔐 Browser sign-in — public OAuth client + PKCE; nothing secret stored
- 🤖 Works with Claude Code, Claude Desktop, and Cursor
Requirements
- Node.js 18+
- A Reel Estate account (a paid plan is required to create/generate/render; free accounts are read-only over the MCP)
- An MCP-compatible client: Claude Code, Claude Desktop, or Cursor
Install
Nothing to clone or build — your MCP client launches it on demand via npx. You
only need Node.js ≥ 18. (Prefer a global command? npm install -g reel-estate-mcp,
then use reel-estate-mcp in place of npx -y reel-estate-mcp below.)
Connect your assistant
Zero config — the bridge connects to Reel Estate and handles browser sign-in for you. There's nothing to set up.
Claude Code
claude mcp add reel-estate -- npx -y reel-estate-mcp
Claude Desktop / Cursor
Add to your client's MCP config (claude_desktop_config.json, ~/.cursor/mcp.json):
{
"mcpServers": {
"reel-estate": {
"command": "npx",
"args": ["-y", "reel-estate-mcp"]
}
}
}
First run — sign in
The first tool call opens your browser to sign in with your Reel Estate account
(Clerk OAuth, authorization code + PKCE). After you approve, the browser returns to
http://localhost:8765/callback, tokens cache under ~/.reel-estate-mcp, and
you're in — you won't log in again until the token expires.
Start with whoami to confirm auth and see your plan, then just ask:
- "List my recent projects."
- "Create a project for 123 Main St and upload the photos in ./listing."
- "Virtually stage the living room photo, generate a drone clip, then render in 9:16."
- "How many credits and exports do I have left?"
To switch users or re-authorize, run logout (revokes server-side + clears
the cache) or delete ~/.reel-estate-mcp.
Tools
All API tools proxy through the backend's /mcp api_request, so the backend's
own authorization and plan rules apply.
| Tool | What it does |
|---|---|
help |
Guided, always-current walkthrough (also the getting_started prompt) |
whoami |
Confirm auth; reports plan, canWrite, and a writeAccess reason. Run first. |
login / logout |
Start browser sign-in / clear the session (server revoke + local cache) |
list_projects · get_project · project_stats |
Browse projects |
list_clips · list_movies · list_voices |
Browse clips, rendered movies, TTS voices |
resolve_address |
Geocode a free-text address → ranked candidates (for real-listing projects) |
add_image_from_file |
Upload a local image into a project (presigned upload — no storage creds) |
generate_clip · get_clip_status |
Animate a photo into a video clip (Runway); poll the job |
edit_image |
AI photo edit — virtual staging, twilight, upscale, seasonal, replace/remove/add, manual |
add_timeline_audio · add_timeline_overlay |
Place voiceover/music/audio or an image/text overlay |
move_timeline_element · reorder_timeline |
Retime / resequence the timeline |
render_movie |
Assemble the timeline into the final listing video |
list_endpoints · api_request |
Discover the API catalog / call any route — the escape hatch |
Paid vs free
Writes and generation over the MCP require a paid plan — free accounts are
read-only (browse projects, clips, and movies). whoami reports this up front;
a blocked write returns 403 MCP_PAID_PLAN_REQUIRED. See plans at
tryreelestate.com.
How auth works
first tool call
│ StreamableHTTP client ──► backend /mcp (401, needs auth)
▼
opens your browser ──► Clerk OAuth (authorization code + PKCE, public client)
│ │
▼ ▼
loopback http://localhost:8765/callback?code=… access + refresh tokens
│ │
└────────────► finishAuth(code) ──► tokens cached ──┘ (~/.reel-estate-mcp)
every later call: callTool("api_request", …) over the authed /mcp connection
- No secrets to distribute — public client + PKCE, browser login per user.
- Prod-capable — uses the same OAuth the backend serves at
/mcp. - The backend is the single auth authority — this bridge never mints tokens.
Troubleshooting
| Problem | Fix |
|---|---|
| No login prompt / "not authenticated" | Run any tool (or login), open the printed URL, and approve. |
| Switch accounts / re-authorize | Run logout, or delete ~/.reel-estate-mcp. |
Writes blocked (403 MCP_PAID_PLAN_REQUIRED) |
Creating/generating/rendering needs a paid plan; free accounts are read-only. See tryreelestate.com. |
Architecture
src/config.ts— validated env (loaded from this package's folder); derives the/mcpURL, OAuth store dir, callback port, read-only flag, and default client id.src/oauth.ts—OAuthClientProvider: uses the publicclient_id, caches tokens + the PKCE verifier, opens the system browser.src/upstream.ts— the single OAuth'd MCP client connection to/mcp(with the loopback callback server);callTool/callApiRequestproxies.src/api-client.ts—ApiClientovercallApiRequest; enforces read-only.src/catalog.ts— the endpoint catalog surfaced bylist_endpoints.src/tools.ts— tools as plain functions (smoke-testable).src/index.ts— registers the tools as MCP tools over stdio.
Adding a tool
- Add
async function fooBar(api, args)insrc/tools.ts(useapi.get(...)/api.request(...), which proxy through/mcp). - Register it in
src/index.tswith a ZodinputSchema. - Add it to
scripts/smoke.tsif it's a GET.
Everything is already reachable through api_request; convenience tools just make
the common paths first-class.
About Reel Estate
Reel Estate helps real-estate agents and marketers turn ordinary listing photos into professional property videos — AI virtual staging, twilight conversion, motion/drone clips, voiceover narration, and one-click rendering for Instagram, TikTok, and YouTube. This MCP server brings that workflow into your AI assistant. Get started at tryreelestate.com →
Keywords: real estate video generator, AI listing video, virtual staging, MCP server, Model Context Protocol, Claude, Cursor, property video marketing, drone real estate video, twilight photo editing.
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。