Skip to main content
AgentTV makes your video library usable by AI. A person asks ChatGPT, Claude, Gemini or your own site a question. The answer points to your video at the right second, names you, and links back to your page. AI apps do not watch video. They read the text that comes with it. AgentTV builds that text for every video you connect, and serves it in the forms search engines and agents read. Your videos keep playing from your own player and your own CDN, under your own rules.
For AI agents. These docs have an index at /llms.txt, the full text at /llms-full.txt, and a Markdown copy of every page at its path plus .md. See For AI agents.

​
Pick your path

You run a video platform

Brands and media companies with their own player and CDN. Import your catalog by your own ids, put metadata on your pages, and choose how videos play in AI apps.

You build agents and apps

Builders. Make one request, get moments with exact times, and connect an AI app over MCP. No key for public libraries.

You are a creator

Connect your YouTube channel or other accounts, choose where your videos appear, and publish your own app in ChatGPT.

You run a paid membership

Make videos members-only, so only your members can search and ask on your member page, and watch in AI apps. Connect Kajabi.

​
How it fits together

  1. Import. AgentTV reads your video sitemap and your metadata API, and stores each video under your own id. See Import a catalog.
  2. Index. AgentTV builds the transcript, chapters, moments, people, places and rights for each video. See The ARV record.
  3. Publish at crawl time. Your page template fetches one block per video and renders it in your page. Search engines and AI crawlers read it. See Put metadata on your pages.
  4. Answer at inference time. When an AI app answers a question, it reads the ARV record or calls the MCP tools for the exact moment. See Crawl time and inference time.
  5. Play. The moment plays in your own player, or in the AgentTV player from your own stream. See Play in your own player.

​
What AgentTV builds for each video

  • Transcript. Timed sentences of what is said, with speakers.
  • Chapters. Titled sections with a short summary each.
  • Moments. The start and end time that answers a question, with the words and on-screen text behind it.
  • People, places, products and topics. Who and what each video covers, so a question about a person finds the right moment.
  • Rights. Your rules for each video: whether AI apps may search, cite, quote, summarize, play or train on it. Every surface reads the same rules, and every use is logged.

​
A typical brand setup

1

Import your catalog

Give us your video sitemap and read access to your metadata API. Pick the videos to import. See Import a catalog.
2

Put metadata on your pages

Your page template fetches one block per video page and renders it server side. See Put metadata on your pages.
3

Choose how videos play

Play in your own player, or in the AgentTV player from your own HLS stream. See Play in your own player.
4

Connect AI apps and your site

Connect the MCP server, add Ask to your pages, and build rails from chapters. See Play a moment in an AI app, Add Ask to your site and Build a rail from chapters.
5

Set access and measure

Allow or block AI use, then track AI answers before and after. See Control AI access and Track AI visibility.
No partnership with an AI company is needed. AgentTV works through the open formats search engines and AI apps already read.

​
Hosts

​
The open standard

The metadata AgentTV publishes follows the Agent-Ready Video (ARV) standard. ARV is one open record that tells AI what is in a video, who owns it and how it may play. The spec lives at agentreadyvideo.org.