A remote MCP server for your VideoDB account.

Connect Claude, Claude Code, Cursor or any MCP client to VideoDB. Upload media from a URL, index it, search for moments, cut clips and get playable stream links from inside a conversation. The server is hosted at one URL and you sign in with your VideoDB account.

https://mcp.videodb.io/mcp

Works with Claude, Claude Code, Cursor and other MCP clients. Read the docs.

Connect a client

Add https://mcp.videodb.io/mcp as a remote server. On first use a VideoDB page opens; sign in and click Authorize. Full walkthrough in the connection guide.

Claude

Web, desktop and mobile

  1. Open Settings → Connectors → Add custom connector.
  2. Paste the server URL and click Connect.
  3. Sign in to VideoDB in the tab that opens and click Authorize.
https://mcp.videodb.io/mcp

On Team and Enterprise plans an admin adds it under Organization settings → Connectors.

Claude Code

One command, then authorize with /mcp

# add the server
claude mcp add --transport http videodb https://mcp.videodb.io/mcp

# inside Claude Code, authorize
/mcp

Cursor

Settings → Cursor Settings → MCP

{
  "mcpServers": {
    "videodb": { "url": "https://mcp.videodb.io/mcp" }
  }
}

Any MCP client

Anything that supports remote MCP servers

Add the URL as a remote MCP server. To inspect the tools without a client, use MCP Inspector:

npx @modelcontextprotocol/inspector

Then ask: “list my videodb collections”. Your collections come back with IDs and names.

How it works

VideoDB indexes speech and scenes at the moment level, so a client can ask for a specific moment in hours of footage without downloading or re-encoding anything. The full list of what the server can do is in the tool reference.

  1. 01

    Upload

    Point it at a public URL. VideoDB pulls the media into a collection and hands back an ID and a playable stream.

  2. 02

    Index

    Transcribe speech, describe what is on screen, or run analyzers for objects, text, brands and activities. Then build a retrieval-ready index.

  3. 03

    Search and ask

    Find moments by meaning with start and end timestamps, or ask a question and get an answer with sources, across one video or a whole library.

  4. 04

    Cut and stream

    Clip, stitch, subtitle, reframe to vertical, render a timeline. Every result is a stream URL with a matching player link.

Workflows

Describe the outcome and the client picks the right steps. More examples in the workflows guide.

Highlight reel from a long video

“Upload the keynote recording at this link, index the speech, then cut every moment where they demo the product into one reel.”

Ask questions of a recording

“Transcribe m-abc123 and set it up so I can ask questions. Start with: what pricing objections came up?”

Subtitled vertical clip for social

“From m-abc123 take 2:10 to 2:40, crop it to 9:16, burn in subtitles and give me an MP4.”

Search a whole library

“Index every video in “Customer calls”, then find every mention of the renewal discount.”

Live camera with alerts

“Connect rtsp://… as “Lobby cam”, watch for anyone entering after hours, and POST to my webhook when it happens.”

FAQ

What is the VideoDB MCP Server?
A hosted remote MCP server at https://mcp.videodb.io/mcp that gives Claude and any other MCP client access to a VideoDB account: upload media, index it, search it, cut clips and get playable stream links from inside a conversation.
Which clients does it work with?
Claude on web, desktop and mobile (as a custom connector), Claude Code, Cursor, MCP Inspector, and any app that supports remote MCP servers. Team and Enterprise Claude admins add it under Organization settings → Connectors.
Do I need a VideoDB account?
Yes. Sign up at console.videodb.io. When your client connects, you sign in to VideoDB once in a browser tab and click Authorize.
Do I have to install anything?
No. The server is hosted. Paste the URL into your client and sign in. Nothing runs on your machine and your API key never goes into a config file. A local stdio package, videodb-director-mcp, exists for clients that only support stdio; see the local server docs.
What does the server collect or store?
Only the arguments of the tool call being run. It reads no conversation content, chat history, memory or files, and it stores nothing. Tool calls are forwarded to the VideoDB API only. Your media and the transcripts, indexes and streams built from it live in your VideoDB account until you delete them. Details in Auth and Privacy.
What can it not do?
It does not generate AI images, video, audio or music, and does not dub or clone voices. It does not upload local files (upload_media takes public URLs), and it does not capture your desktop. The VideoDB SDK and Agent Skills cover those.
Why did a tool return status: processing?
Uploads, indexing runs and renders can outlast a single request. The job keeps running on VideoDB; the response names the read tool to check with in its retry_with field. Ask your client to check with that tool instead of re-issuing the write, which would start a second job.