How to read EXIF, XMP, and IPTC data from images with AI Leave feedback

Images carry the richest metadata of anything on a normal disk: the camera and its settings, the moment of capture, the editing history, and often the exact location. All of it is one tool call away — and none of it leaves your machine.

Note

The commands and config snippets on this page are for the .NET build of the server — the only platform available today. Installation and client setup: MCP server for .NET. Other platforms will expose the same tools with their own launch command; everything else on this page applies unchanged.

What EXIF data do these photos have? Include camera, date taken, and GPS if present.

read_metadata returns the packages the file carries — EXIF, XMP, IPTC — and the agent reads the answer out of them.

Which of these photos were taken with the same camera? Sort these images by date taken, not by file date. Do any of these have GPS coordinates? Where? Does the XMP show which software edited this?

The second one is quietly useful: file timestamps get rewritten by copying and syncing; EXIF date-taken does not.

In evaluation mode GPS data and thumbnails are unavailable — the read simply does not include them. An unlicensed check that reports “no location data” is not evidence of anything. Before concluding a photo is safe to publish:

What is the license status of the metadata server?

get_license_status; see Licensing.

Strip EXIF from every image in this folder and confirm the copies are clean.

remove_metadata writes cleaned copies; a second read on the copy confirms. Remember the originals are still on disk with everything intact — publish the copies. Full workflow: Strip metadata before sharing.

Sending a photo to a cloud service to ask what it contains sends the photo — including the location you were trying to check for. Here the image is read in place by a local process, and only the answer travels; with a locally-hosted model, not even that. See On-premise architecture.