Video Understanding MCP for VS Code
io.github.granitebps/video-understanding-mcp
Let AI agents watch videos and answer questions about them through any OpenAI-compatible endpoint.
client:VS Code
transport:stdio
runtime:npm
Install Video Understanding MCP in VS Code
.vscode/mcp.json
{
"inputs": [
{
"type": "promptString",
"id": "video-base-url",
"description": "VIDEO_BASE_URL",
"password": true
},
{
"type": "promptString",
"id": "video-model",
"description": "VIDEO_MODEL",
"password": true
},
{
"type": "promptString",
"id": "video-api-key",
"description": "VIDEO_API_KEY",
"password": true
},
{
"type": "promptString",
"id": "video-extra-headers",
"description": "VIDEO_EXTRA_HEADERS",
"password": true
},
{
"type": "promptString",
"id": "video-part-format",
"description": "VIDEO_PART_FORMAT",
"password": true
},
{
"type": "promptString",
"id": "video-max-duration-seconds",
"description": "VIDEO_MAX_DURATION_SECONDS",
"password": true
},
{
"type": "promptString",
"id": "video-max-upload-mb",
"description": "VIDEO_MAX_UPLOAD_MB",
"password": true
},
{
"type": "promptString",
"id": "video-allowed-dirs",
"description": "VIDEO_ALLOWED_DIRS",
"password": true
},
{
"type": "promptString",
"id": "video-verify-media",
"description": "VIDEO_VERIFY_MEDIA",
"password": true
}
],
"servers": {
"video-understanding-mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@granitebps/video-understanding-mcp"
],
"env": {
"VIDEO_BASE_URL": "${input:video-base-url}",
"VIDEO_MODEL": "${input:video-model}",
"VIDEO_API_KEY": "${input:video-api-key}",
"VIDEO_EXTRA_HEADERS": "${input:video-extra-headers}",
"VIDEO_PART_FORMAT": "${input:video-part-format}",
"VIDEO_MAX_DURATION_SECONDS": "${input:video-max-duration-seconds}",
"VIDEO_MAX_UPLOAD_MB": "${input:video-max-upload-mb}",
"VIDEO_ALLOWED_DIRS": "${input:video-allowed-dirs}",
"VIDEO_VERIFY_MEDIA": "${input:video-verify-media}"
}
}
}
}
VS Code prompts for each input the first time the server starts, so no secret is written to the file. VS Code docs