MCP App Store
by KeenEthics
Oracut icon

Oracut MCP Server for ChatGPT

By Oracut

Overview

Oracut turns your videos into short clips – for anyone who wants to get reels and shorts out of talking-head videos, webinars, lives or interviews without an editing program. How it works: - Upload: 1 to 5 videos that become one clip – new ones or videos you uploaded before (15 seconds to 45 minutes each, together up to 45 minutes, up to 6 GB, MP4, MOV or WebM). Optionally say or type what the clip should focus on. - Target length: choose 15 to 180 seconds. If the videos are longer, or you upload several takes, Oracut first works out the core message, uses the best take of every part and puts them in the best order (hook, core message, clear ending). - Cut: pauses, "ähm", false starts and sentences said twice are removed automatically; landscape videos become 9:16 with the frame following the speaker, or stay landscape on a blurred or coloured background, optionally with a headline at the top. - Cut pace and hook: slow, medium or fast cuts with soft transitions instead of hard jumps, and optionally a short, catchy hook from the clip at the very start. - Captions: three caption templates with the spoken word highlighted; save your colours as presets. - Editor: correct words, change colours, position, headline and background, take sentences out or add them, move single cuts in the fine cut, change pace, transitions and hook, or have the clip cut again with new wishes, then download the MP4 (720p up to 4K; some accounts add a visible Oracut watermark) and the captions as SRT or VTT. Limitations: Oracut does not publish anything; you download the clip and post it yourself. Raw videos are deleted after 7 to 14 days, finished clips after 7 to 60 days. The spoken language is detected automatically (German, English and many more).

How to connect

  1. ChatGPT: Open the Oracut listing in the ChatGPT app directory and choose Connect. ChatGPT listing Oracut exposes 24 tools here.

Tools

complete_video_upload

ChatGPT
Used by the upload window after all parts of a video were uploaded: joins the parts (partNumber + ETag of each part). Then create_reel.

create_reel

ChatGPT
Use this after the videos were uploaded (start_video_upload + complete_video_upload) or with videos uploaded earlier (list_videos). Makes ONE 9:16 clip from one to 5 videos in which someone speaks (assetIds in any order – takes of the same text or parts that belong together; each 15 s to 45 min, together at most 45 min) and reserves the credits (1 per started minute of all videos together, at least 2 per clip; resolution 2160p (Pro) +2 per started minute of target length and clip):if the videos are longer than targetSeconds (default 60, 15–180 s) or there are several videos, the AI first plans the clip: it works out the core message, finds every take of each part in all videos, uses the best one (fluent, clear, good sound) and puts the parts in the order that tells the message best – no matter in which order the videos came – then reviews the script (opening, no repeats, clear ending), following focus if given; then long pauses, filler words (äh, ähm), false starts and sentences said twice are cut out, in the chosen pace (default medium), with soft transitions at the cuts (default on). hook true: the AI picks a short, catchy line of the clip (no set length, never the line the clip opens with) and shows it once more at the very start (default on; hook false switches it off). Adds captions (subtitles from the spoken words, editable; captions false switches them off). Landscape videos: layout reframe (default) makes a full 9:16 picture whose frame follows the speaker; layout fit keeps the whole landscape video in the middle of the 9:16 picture with a background above and below (background: blur = the video blurred, or a colour #RRGGBB). Fit only works when ALL videos are landscape; portrait or mixed videos are always reframed. headline: optional short title shown at the top for the whole clip (up to 90 characters, up to 3 lines; changeable later with update_captions). clips (Pro plan, up to 5): makes that many DIFFERENT clips from the same videos – each with its own core message and its own passages, no passage twice – for the minutes of the videos once, at least 2 credits per clip (fewer clips when the videos are too short for that many clips of the target length); returns the first clip with reelIds of all of them. Renders in the background and returns reel_id at once; poll get_reel until status is ready (usually a few minutes). Calling it again within two minutes with the same videos and settings returns the clip already started; later the same videos can be used for a new clip.

delete_color_preset

ChatGPT
Use this to delete one saved caption colour preset for good (id from list_color_presets). Clips already coloured with it keep their colours. If it was the default, new clips get the standard colours.

delete_reel

ChatGPT
Used by the window when the user deletes one of their clips: the clip, its video files and its transcript are removed at once. Not while the clip is being made.

delete_video

ChatGPT
Used by the window when the user deletes one of their uploaded videos: the file is removed at once and its storage is free again. Clips made from it stay, but can no longer be cut anew.

export_content

ChatGPT
Use this to create a download of a finished clip (only when get_reel says ready): mp4 (the video), srt or vtt (the captions as a subtitle file). The signed link expires after 1 hour; give it to the user as a link.

get_account

ChatGPT
Use this to show the user their plan and remaining credits (1 credit per minute of source video, at least 2 per clip, 4K extra), the export resolution and how long files are kept.

get_captions

ChatGPT
Use this to read the captions (subtitles burned into the video) of a reel: blocks with id, time and text, plus the style. Needed before correcting a misheard word with update_captions.

get_cut_sources

ChatGPT
Used by the window for the fine cut preview: signed links to the original videos of a clip and their place on the clip timeline.

get_duplicate_takes

ChatGPT
Use this when a reel has sentences the speaker said more than once. Returns each group with its takes (text, sentence id, length) and which take is kept. Pick the better take yourself or ask the user briefly, then call update_reel with keepTakes.

get_reel

ChatGPT
Use this to check a clip: status, length before/after the cut, removed passages with reason, the sentences with ids (for update_reel), resolution, the credits of a change, and a download link once ready.

list_color_presets

ChatGPT
Use this to list the user's saved caption colour presets (text, accent of the spoken word, box). The default preset colours every new clip.

list_reels

ChatGPT
Use this to find the user's clips (newest first) with title, status and length, optionally filtered by status (e.g. ready).

list_videos

ChatGPT
Use this to show the videos the user uploaded before and that are still kept (raw videos are deleted after the retention period of the plan). Their assetIds can be passed to create_reel again – alone or combined with other or newly uploaded videos, up to 5 per clip.

open_oracut

ChatGPT
Use this to open the Oracut window: upload 1–5 videos for one clip or reuse earlier uploads (choose target length, captions, cut pace, transitions, hook, focus), see all clips with their progress, edit, cut again or download a clip, and manage caption colour presets. Optional id (a reel id) opens that clip in the editor.

recut_reel

ChatGPT
Use this when the user wants the clip cut again differently (e.g. 'focus more on the pricing part', 'make it calmer', typed or spoken in the window): the AI picks passage and hook anew from the same videos, following instructions, optionally with a new pace, transitions, hook or target length. It costs 2 credits, a 4K clip plus its 4K surcharge (get_reel recutCredits) – tell the user and get their OK first. Only when the clip is ready. Runs in the background; poll get_reel until ready. Small changes (remove a sentence, other pace) are cheaper with update_reel (get_reel editCredits).

render_reel_final

ChatGPT
Use this to export the final video in full resolution after changes: after a changed cut (720p preview) the export is included; after only caption changes it costs 1 credit, a 4K clip plus its 4K surcharge – get_reel editCredits (get the user's OK first). It replaces the previous final video of the clip. Does nothing if the final video is already up to date, so calling it again costs nothing. Runs in the background; poll get_reel until status is ready.

save_color_preset

ChatGPT
Use this to save a caption colour preset (max. 12): name, textColor, highlight (colour of the spoken word, null = none), bgColor (box, null = none), bgOpacity 0–1. With id it overwrites that preset; without id it adds a new one. isDefault true makes it the default for new clips. To use it on a clip, call update_captions with these colours.

set_default_color_preset

ChatGPT
Use this to choose which saved caption colour preset colours every new clip (id from list_color_presets), or null for no default (standard colours). Existing clips do not change.

show_reel_editor

ChatGPT
Use this to open the Oracut window on the editor for a clip: video preview with captions, correct caption text, change font, colours (and saved colour presets) and position, adjust the cut (sentences, pace, transitions, hook) or have it cut again, then render the final video and download it.

show_video_upload

ChatGPT
Use this to open the Oracut window on the upload when the user wants to turn videos into a clip. The user picks 1 to 5 videos that become one clip (max 45 min / 6 GB each, together max 45 min), the target length, captions, cut pace, transitions, hook and an optional focus there; the window uploads them straight to storage, starts the clip and shows the progress. Videos attached in the chat cannot be used – always use this window.

start_video_upload

ChatGPT
Used by the upload window: checks credits and limits (max 45 min and 6 GB per video, up to 5 files) and returns signed multipart upload links straight to storage (valid 1 hour). Each part is PUT to its URL; then complete_video_upload with the ETags, then create_reel with all assetIds of the clip. Each call creates new, unused upload slots.

update_captions

ChatGPT
Use this to correct caption text (blocks: id + text, e.g. a misheard word), switch captions on or off (enabled), pick a template (schlicht, verspielt, modern), or change the look: style.font (montserrat, anton, lora), textColor, bgColor (null = no background), bgOpacity 0–1, position {x,y} as share of the frame (centre of the block), size 0.7–1.8, highlight (colour of the word being spoken, null = off); per block override colours/position. headline: title at the top of the clip (null or "" removes it); background (only clips made with layout fit): blur or #RRGGBB. Overwrites the previous text and look of the changed blocks. Only saves – call render_reel_final afterwards so the video shows the change.

update_reel

ChatGPT
Use this to change the cut of a reel and render it again from the original video: removeSentences / restoreSentences take sentence ids from get_reel, keepTakes chooses the take of a duplicate group (from get_duplicate_takes), title renames it. For a clip with a selection (kind short: shortened to its target length, or made from several videos), selectSentences sets the complete new selection in PLAY order – it may jump between videos and back, e.g. the best take of each part (sentence ids from get_reel; keep it within the target length); restoreSentences adds sentences next to their neighbour; the plan in get_reel lists the takes of every part. cutEdges is the fine cut: it moves edges of the kept ranges (get_reel ranges, by index; start/end in seconds on the same timeline, at most 3 s from the automatic edge aiStart/aiEnd) – e.g. a cut that clips a word: end of range 2 +0.2. Send it alone; the first fine cut of a clip is free, later ones cost like any change. pace (slow, medium, fast), transitions (on/off) and hook (true: the AI picks a short, catchy line for the start – never the one the clip opens with, false: removes it) change the cut as well; hookSentence makes that sentence (id from get_reel, in the clip) the hook – the user's choice, no AI. A changed cut costs 1 credit, a 4K clip plus its 4K surcharge – get_reel editCredits (get the user's OK first; the first fine cut of a clip is free) and renders a 720p preview; after checking it, render_reel_final exports it once in full resolution at no extra cost. Poll get_reel until ready. For a completely new cut with other wishes use recut_reel.

Capabilities

InteractiveWrites

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/oracut"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/oracut)