upload_videoLABSCO SUMMARY
Find the moment, not the video.
Filename and title search stops at the file; this goes inside it, so 'the bit where someone opens the gift' is a query rather than an afternoon of scrubbing. Because import covers creators and hashtags as well as your own uploads, the same search works over material you did not shoot.
What it is
A video understanding layer built on memories.ai. Videos are uploaded or imported, indexed, then searched by what happens in them; the same connection carries captioning, transcription and a persistent memory store.
What you get
- Videos uploaded from a URL and indexed, with status, listing and deletion
- Semantic search across your own library, and across public TikTok, YouTube and Instagram
- Audio search, for finding the moment something was said
- Multi-turn questions about one video or the whole collection
- Import by creator or by hashtag, plus direct import from a URL
- Captioning of any video or image from a URL, without uploading it first
- Visual or spoken transcription, and a memory store you can add to, list and search
Requirements
A memories.ai account and its API key; the free tier carries 100 credits a month. The server is built from source and launched by your client with the key in its environment.
Setup effort
Build from source — git clone https://github.com/kennyzheng-builds/videoseek-mcp.git
USE CASES
01
Finding one moment in a library you cannot watch end to end
Describe what happens and search your indexed videos by meaning, then confirm against the transcription rather than scrubbing a timeline.
02
Researching what a creator is doing
Import a creator's recent work, wait for indexing, then ask questions across the whole set instead of watching twenty clips.
03
Tracking a trend across platforms
Search public TikTok, YouTube and Instagram by meaning, then pull in by hashtag what is worth keeping.
04
Finding where something was said
Audio search locates the spoken line inside a video, which is the fast path when you remember a phrase but not which recording it was in.
05
Describing a clip without adding it to your library
Caption a video or image straight from its URL when you need to know what is in it and do not want it indexed or stored.
06
Keeping what you learned across sessions
Add findings to the memory store and search them later, so a research pass through a library survives into the next conversation.
18 tools
Listed in project docs. Your client may show a different list.
list_videosget_video_statusdelete_videosget_transcriptionsearch_videossearch_publicsearch_audiochat_with_videochat_personaladd_memorysearch_memorieslist_memoriescaption_videocaption_imageimport_from_urlimport_by_hashtagimport_by_creatorBEFORE YOU INSTALL
What to know first
Tools you get
18 tools named in the project's documentation.
Requires
MEMORIES_API_KEY in your environment
Where it runs
All 3 documented clients run the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
git clone https://github.com/kennyzheng-builds/videoseek-mcp.git
cd videoseek-mcp
npm install --include=dev
npm run buildMCP Clients
Claude Desktop, Cursor
JSON
{
"mcpServers": {
"videoseek": {
"command": "node",
"args": [
"<YOUR_PATH_TO_VIDEOSEEK_MCP>/build/index.js"
],
"env": {
"MEMORIES_API_KEY": "<YOUR_MEMORIES_API_KEY>"
}
}
}
}Claude Code
CLAUDE CODE · TERMINAL
claude mcp add videoseek -- node <YOUR_PATH_TO_VIDEOSEEK_MCP>/build/index.js
