Playback.tv
Create

MCP connector

Use Playback from Claude or ChatGPT

Add one address to your assistant, then ask it for a moment: “loop 1:23 to 1:35 of this video” or “make a Cut of the three best exchanges in this episode”. It answers with a playback.tv link that opens on exactly that part. No account, nothing to install.

Connector address

https://playback.tv/api/mcp

make_loop_link

One YouTube video, a start and an end, an optional speed and note. Returns a loop link that plays that part on repeat.

make_cut_link

2 to 12 moments from up to six videos, each with its own note. Returns a Cut that plays them in order.

Claude

  1. 1On claude.ai or the Claude desktop app, open the connector settings and choose to add a custom connector.
  2. 2Name it Playback and paste the address above. There is no client ID and no sign-in to set up.
  3. 3Turn it on in a chat and ask for a moment. Claude calls the tool and hands you the link.

In Claude Code: claude mcp add --transport http playback https://playback.tv/api/mcp

ChatGPT

  1. 1In ChatGPT’s settings, turn on developer mode, where custom MCP connectors live.
  2. 2Create a connector with the address above and choose no authentication.
  3. 3Pick the connector in a chat and ask for a moment.

Any other MCP client

It is a stateless Streamable HTTP server: JSON-RPC over POST, no session, no authentication. Most clients take a config like this:

{ "mcpServers": { "playback": { "url": "https://playback.tv/api/mcp" } } }

Questions

Does it need my Playback account?
No. Both tools only make links, so there is nothing to sign in to and nothing is saved. Open the link and save it from the page if you want to keep it under your name.
What does it send to YouTube?
Only the video's address, to YouTube's public oEmbed endpoint, to read the title and channel and to check that the video can be embedded. Nothing is stored. The video itself is never fetched: the link plays it in YouTube's own player.
Can it find a moment for me?
It works from the times you or your assistant give it. If the assistant can read the video's transcript or chapters, it can pick the times from those and pass them along.