Skip to content
FilmScribe
Workflow

What is MCP?

MCP (Model Context Protocol) is an open standard that connects an AI assistant to an outside service. With an MCP connector, ChatGPT, Claude, Gemini, Cursor or VS Code can look up data and take actions in that service during the conversation, always with your permission and within what the connection allows.

Also called: Model Context Protocol, MCP connector, custom connector

See the connector Try it with my video
Example

A request through the connector

You ask in plain language. The assistant queries the service through the connector and answers with real data, like exact timestamps.

You Request What were the three strongest moments of episode 42? Give me the time of each one.
Assistant Answer From the episode's hooks: 00:18:42, the rain story; 00:31:05, the audio mistake; 00:52:10, the closing advice.
Why it matters

Why MCP matters

Without a connector, you copy passages and paste them into the chat, and the assistant answers about a slice of the material. With a connector, it reads the source directly, in full, and cites the right data.

You stay in control: the connection only exists if you authorize it, you decide whether it only reads or can also make changes, and you can revoke it whenever you want.

In FilmScribe

The FilmScribe MCP connector

Our connector links your account to ChatGPT, Claude, Gemini, Cursor or VS Code. By default access is read-only: the assistant looks up transcripts, chapters, emotions, hooks, scripts and clips, and looking things up does not use hours. If you allow it when connecting, it also organizes folders, generates chapters and scripts, translates and renders clip videos. Deleting and using hours always wait for your yes. It comes with Studio and Agency.

AI connector Ask your assistant about your recordings and get answers with the right timestamps, links and timelines. See the connector →
In practice

Using an MCP connector safely

  1. 01

    Grant the least access that works

    Start with read-only access and add write permissions only for tasks you actually want the assistant to do. You can always widen access later.

  2. 02

    Connect only services you trust

    A connector sees what the assistant sends it and returns text the assistant will act on. Add connectors only from services you trust, using the address from their official site.

  3. 03

    Read before you approve

    When the assistant asks to run an action, read what will change: which file, which folder, how many items. Approving out of habit is how unwanted changes slip through.

  4. 04

    Prune old connections

    Remove connectors you no longer use, both in the assistant and in the service. Fewer active connections means less to watch and less exposure if an account is compromised.

Related terms

See all terms →
Transcription The speech in audio or video written down as text, either verbatim, with every hesitation, or clean and easy to read. XML and FCPXML Files that describe a whole timeline, with segments, tracks and timing, so a video editor can open it. YouTube chapters A video split by topic, created from the list of timestamps and titles you paste into the description. Hook The line or scene at the start of a video that makes someone stop scrolling and want to see the rest.

Frequently asked questions

Which assistants support MCP?

The ones that accept custom connectors, including ChatGPT, Claude, Gemini, Cursor and VS Code. Each assistant decides which of its plans include that option.

Can the assistant change my files?

Only if you allow it when connecting. By default, the FilmScribe connector is read-only. With permission, deleting and using hours still ask for your confirmation.

How do I revoke an assistant's access?

In FilmScribe, go to Profile > Connections and revoke the connection. It also expires on its own after 45 days without use or when you change your password or email.

Captions, clips and paper edits. All in one place.

Upload a video or paste a YouTube link. We transcribe every word with its timing, and you take the result to social media or to your editor.

Try it with my video
See the result on your own recording before adding a card.