inspect_media MCP tool

MCP name inspect_media dispatches to renderer inspectMedia.

Does not use hosted quota. Local editor operation.

Look at project media. action inspect views an asset: images and audio return the media, video and screen recordings return timestamped contact sheets (optional startTime and endTime as second strings), components return source. action captureStill saves a still from an asset. action captureFaces saves detected faces. Editor must be open. Does not use hosted quota. Local editor operation.

Advertised MCP input

This tool is registered with passthrough. The fields below are what listTools advertises. The renderer still validates the child fields for the chosen action or kind. Extra advertised-only values are documented under Reachable behavior.

Field	Type	Required	Constraints	Default	Description
action	enum("inspect", "captureStill", "captureFaces")	yes	—	—	action for inspect_media
assetId	string	no	minLength 1	—	Required for inspect, captureStill, and captureFaces.
timecode	number	no	min 0	—	Required when action is captureStill. Asset time in seconds. 0 is valid.
startTime	string	no	—	—	Optional inspect range start in seconds as a string, for example "2.5". Pair with endTime.
endTime	string	no	—	—	Optional inspect range end in seconds as a string, for example "10". Pair with startTime.
resultStart	integer	no	min 0; max 9007199254740991	0	First character offset to return. Use 0 for the first call.
resultEnd	integer	no	min 1; max 9007199254740991	8000	Exclusive end offset. resultEnd - resultStart must be at most 8000. Increment resultStart by 8000 to page through large results.

Reachable behavior

Advertised action values match the renderer children: inspect, captureStill, captureFaces.

Renderer child fields

These are the authoritative fields for each action. Send them on the same MCP call. Grouped MCP tools use passthrough, so these fields are not all listed in the advertised schema.

action=inspect

Renderer tool: inspectAsset.

Load an asset by ID so you can see/hear it or read its description. For image or audio: returns the media for viewing. For SVG images: returns a sanitized PNG preview when possible and sanitized SVG source code. For 3D model assets (.glb/.gltf): returns a structural summary of the model (scenes, meshes, materials, animations). For component assets: returns name, class, inputs, and full source code (no playable URL). For video or screen recording assets: returns contact-sheet stills with timestamps, covering the full asset or startTime–endTime.

Field	Type	Required	Constraints	Default	Description
assetId	string	yes	minLength 1	—	—
startTime	string	no	pattern ^\d+(?:\.\d+)?$	—	Start time in seconds, for example '2.5' or '10'
endTime	string	no	pattern ^\d+(?:\.\d+)?$	—	End time in seconds, for example '5.0' or '15'

action=captureStill

Renderer tool: captureAsset.

Bookmark a timeline moment for an uploaded asset using its asset ID and capture an exact still.

Field	Type	Required	Constraints	Default	Description
assetId	string	yes	minLength 1	—	—
timecode	number	yes	min 0	—	—
notes	string	no	—	—	—

action=captureFaces

Renderer tool: captureFaces.

Capture frames from a video at face detection timestamps with bounding boxes drawn around detected faces.

Field	Type	Required	Constraints	Default	Description
assetId	string	yes	minLength 1	—	—
faceIndex	number	no	min 0	—	—
notes	string	no	—	—	—

Result window

Every local MCP tool accepts resultStart and resultEnd so large payloads can be paged.

• resultStart — first character offset. Default 0.
• resultEnd — exclusive end offset. Default 8000.
• resultEnd - resultStart must be at most 8000 characters.
• Increment resultStart by 8000 to read the next window.