Captions Space

AI video editing for Reels and Shorts: Claude, ChatGPT or Cursor add animated subtitles, text, emoji, effects, zooms, b-roll and sounds.

Documentation

MCP server for video editing: Claude and ChatGPT edit Reels and Shorts

Connect your AI agent to Captions Space. Describe the video you want and share examples you like — the agent directs, our studio transcribes, renders and shows it the result. You watch the edit live in the editor.

Server URL

https://api.captions.space/mcp

Transport

Streamable HTTP

Authorization

OAuth (claude.ai, ChatGPT) or an API key in the Authorization header

Clients

Claude, ChatGPT, Claude Code, Cursor, Codex and other MCP clients

Tools

21 tools: speech, scenes, frames, edit, assets, preview, export

MCP (Model Context Protocol) is the standard way for AI assistants to use outside tools. Our MCP server turns Captions Space into an editing studio for your agent: it gives the model what it cannot do itself — hear the speech, see the frames, know exact timings — and renders whatever the model decides.

The agent stays the director: it plans hooks, accents, emoji, zooms and b-roll from your brief. The studio computes pixels and milliseconds, keeps subtitles off faces and renders the video exactly like the final file.

Describe it

Tell your agent what the video should feel like, what to stress, which examples you like.

The agent reads the video

Transcript with word timings, cuts, faces and objects, frame sheets — everything a director needs.

It directs

Style, accents, hooks, emoji, effects, zooms, b-roll, sounds. It brings its own assets and saves the edit.

Checks and iterates

Renders fragments exactly like the final file, looks at them and fixes. You see every change live in the editor.

Animated subtitles

Pick a template, pace and highlighted words; fix any word of the transcript.

Text, emoji and effects

Hooks and titles, emoji next to the speaker, confetti, fire, transitions on cuts.

Zooms and b-roll

Punch-ins on the strongest word, slow pushes, cutaways from its own clips and images.

Sounds and music

Whooshes and hits from our library, your own music track at the right volume.

Its own assets

Images, GIFs, clips and music the agent found or generated elsewhere go straight into the edit.

Translation

Subtitles in 37 languages with the timing kept.

https://api.captions.space/mcp

claude.ai and ChatGPT

Add a custom connector with the server URL below and sign in to Captions Space when asked — no key needed.

Claude Code

claude mcp add --transport http captions-space https://api.captions.space/mcp \
  --header "Authorization: Bearer cs_live_…"

Cursor and other clients with mcp.json

Use a key from Developers:

{
  "mcpServers": {
    "captions-space": {
      "url": "https://api.captions.space/mcp",
      "headers": {
        "Authorization": "Bearer cs_live_…"
      }
    }
  }
}

Example request to your agent

Edit my video into a 30-second Reel: put the strongest line first, bold yellow subtitles, an emoji on every joke, a punch-in zoom when I say the price, and keep the bottom free for Instagram's UI. Show me frames before exporting.

ToolWhat it does
editing_guideHow to edit: workflow, the edit document, ready recipes
list_videos / get_videoProjects, status and the editor link
create_video / start_processingA new video by link or by file
wait_for_videoWait for processing to finish
get_transcriptWords with ids and timings, phrases
get_scenesCuts, who and what is in frame, emotion
get_framesA sheet of source frames — the agent's eyes
get_edits / save_editsThe whole edit, versioned
upload_asset_from_url / request_asset_uploadYour b-roll, images, GIFs, music and sounds
add_emojiEmoji stickers
render_preview / wait_for_jobRender a fragment exactly like the file and look at it
list_styles / get_catalogSubtitle templates, effects, transitions, sounds
export_video / list_exportsMP4, MOV, WebM, SRT, VTT, ASS
account_infoPlan and credits left

It is a set of tools an AI assistant can call through the Model Context Protocol. Captions Space gives the assistant the transcript, scenes and frames of your video and renders the edit it decides on: subtitles, text, emoji, effects, zooms, b-roll and sounds.