Which media library can AI agents upload to and search through an API?

MediaTrove can. It is a photo and video library with a plain HTTP and JSON API, so an AI agent can upload files, search them in plain words, fetch them and tag them, in the same library your people use.

Last updated 28 September 2026.

What can an agent do through the API?

An agent can upload, search, find similar items, fetch files and video keyframes, add and remove tags, add notes, make and revoke share links, and delete files.

Tags an agent adds through the API are stored as human tags, not AI tags.

How does an agent sign in?

Each agent uses its own API key, sent as “Authorization: Bearer” followed by the key. A person signed in to the library makes the key in Settings, and can revoke it at any time.

• A key is shown once, when it is made. MediaTrove stores only a hash of it.
• A key works only at its own library’s address, such as https://your-library.mediatrove.ai.
• A key can’t make or revoke keys. Only people can.
• Settings shows each key’s name and when it was last used.

What does a search request look like?

The agent sends a plain-words query and gets back a list of results. A video result carries the moment, with its start and end time in seconds, its description, its transcript and its tags.

GET /api/search?q=dancers+at+sunset
Host: your-library.mediatrove.ai
Authorization: Bearer mt_…

# the shape of the response, shortened
{ "query": "dancers at sunset",
  "results": [ {
    "file": { "id": "…", "kind": "video",
              "name": "harbour-sunset-dance.mov" },
    "moment": { "id": …, "start": 60, "end": 63.3,
                "description": "…", "transcript": "…",
                "tags": [ { "tag": "…", "source": "ai" } ] },
    "score": …,
    "url": "https://your-library.mediatrove.ai/#file=…&t=60"
  } ] }

score runs from 0 to 1 and says how well the result matches. Results under 0.3 are not returned. url opens the library at that file and moment, so an agent can hand a person a link.

How does an agent upload a file?

For a file under 90 MB, the agent sends the bytes in one PUT request to /api/upload with the file name. Larger files go up in parts.

The file is stored and can be found by name at once. Photos and PDFs are usually described and tagged within a minute. Videos take longer. The agent can ask for the file again to see whether processing has finished (ai_status and proc_status).

A zip is unpacked, and every file inside is filed on its own.

What can’t the API do yet?

There are no read-only or limited keys: a key can do anything a person can do in the library, except manage keys. That includes deleting files, and a delete is permanent.

• There is no SDK and no MCP server. It is plain HTTP and JSON.
• There are no webhooks. An agent asks again to see whether a file has been processed.
• The API is an early version and may change.
• Pricing has not been published yet. It is free to try, with no card.

Where is the full API guide?

Every library serves its own guide at /api/help, as plain text an agent can read. It needs a key or a signed-in person, so set up a library first.

To start: set up a library, open Settings, make a key, and give the key and your library’s address to your agent.

About MediaTrove

MediaTrove is a photo and video library for individuals and small teams. It is a web app at mediatrove.ai, made by 123TEQ, LLC, and first released in 2026. It is not connected to any other product or company with a similar name. Questions: hello@mediatrove.ai.

Free to try, no card needed. Pricing has not been published yet.