BatchEdits MCP use case

    Add captions to a folder of videos with your AI agent

    Tell Claude, ChatGPT or Codex which videos to caption and in what style. The agent uploads them to BatchEdits, applies the style, waits for processing and hands you a download link (or the files) for every captioned video.

    What you say to the agent

    Plain sentences work. A few examples:

    • Claude Code or Codex (terminal)

      “Caption every video in ./raw with my "Captions" style and save the results to ./captioned. Ask me before you spend credits.”

    • Claude or ChatGPT (chat)

      “Create a BatchEdits style called "Captions" with white word-by-word captions at the bottom, then open the uploader so I can add this week's videos.”

    • Any agent

      “Which of my BatchEdits videos are still processing? Give me download links for the finished ones.”

    Set it up

    Chat apps connect to the MCP server. Agents that can read your disk (Claude Code, Codex) can also use the CLI, which uploads a local folder in one command.

    1. Connect BatchEdits to your agent

      Claude (web, desktop and mobile)

      BatchEdits is in Claude's connector directory: open BatchEdits in Claude, click Connect and sign in. Or open Settings → Connectors, choose Add custom connector, enter https://batchedits.com/api/mcp and sign in with your BatchEdits account.

      ChatGPT

      Open Settings → Apps & Connectors, create a new app/connector (developer mode may need to be enabled), enter https://batchedits.com/api/mcp, choose OAuth and sign in.

      Claude Code (MCP)

      terminal
      claude mcp add --transport http batchedits https://batchedits.com/api/mcp
      
      # then run /mcp inside Claude Code to sign in

      Claude Code, Codex or any terminal agent (CLI)

      terminal
      npx batchedits login
      npx batchedits run ./raw --style "Captions" --out ./captioned

      login opens your browser to approve the CLI; the token is stored locally and never printed, so the agent never sees it. See the CLI docs for every command.

    2. Pick or create a caption style

      Ask the agent to list your styles, or to create one with captions turned on (font, color, position, words per line). With the CLI: batchedits style create "Captions" --captions. Add --silence or --crop 9:16 to cut silences or crop vertical in the same pass.

    3. Hand over the videos

      In Claude or ChatGPT the agent opens the BatchEdits uploader in the chat, where you choose the files (in ChatGPT you can also attach a video to the message). Terminal agents run the CLI on the folder: it uploads up to 50 files per run, so a bigger folder is split into runs of 50.

    4. Confirm and let it run

      Each video uses one credit. The agent should ask before spending them (the CLI refuses to process without --yes when there's no terminal to confirm in).

    5. Collect the captioned videos

      The CLI waits and downloads every finished video to the output folder. Over MCP, the agent checks the status and returns temporary download links for each completed video.

    What you get

    • Every video in the folder with word-by-word captions burned in, in the same style.
    • Captions in the language that's spoken: BatchEdits transcribes with OpenAI's Whisper model and supports 99 caption languages. Accuracy depends on the audio quality and the language.
    • The same resolution and aspect ratio as each upload, unless your style crops it.
    • The videos also stay in your BatchEdits dashboard, so you can review or download them there later.

    What it costs

    The MCP server and CLI are included in every account and use the same credits as the web app: 10 videos in the 7-day trial, then 30 a month on Creator Starter. To caption a single video without an account, use the free caption generator.

    Ready to edit from AI?

    Claude Desktop users connect in one click. Everyone else just needs an API key.

    Works with Claude Code, Claude Desktop, Cursor, Windsurf, ChatGPT, and any MCP-compatible client.