Video editing agent for your AI assistant

Connect SendShort once and any assistant — Claude, Cursor, OpenClaw, Hermes, ChatGPT — can import a YouTube link, cut the best moments, caption them and hand back post-ready shorts.

Terminal

# Add the SendShort connector

npx sendshort-agent connect

# Set your API key

export SENDSHORT_API_KEY=ss_live_•••••

# Check the assistant can see it

sendshort tools:list

# Make your first shorts

sendshort clips:create \

--source "https://youtu.be/xxxxxxxx" \

--clips 5 --ratio 9:16 --captions

Built for automation

Everything the agent can reach

Assistant ready

Every tool returns structured JSON, so Claude, Cursor or ChatGPT can plan, call and verify a render on its own.

Long video → shorts

Point it at a YouTube URL or a file and it scores retention, cuts the moments and reframes to 9:16.

Captions & B-roll

Word-level captions at 98%+ accuracy, brand-kit styling, automatic B-roll and zooms on every clip.

MCP connector

One connector, any assistant. Claude, Cursor, OpenClaw, Hermes, ChatGPT and Codex are all tested.

Scheduled runs

Give the agent a cron heartbeat and it clips new episodes the morning they land — no prompt needed.

Same engine as the API

The agent calls the exact endpoints your backend can call. No second pipeline, no drift in output.

How it works

Discover the workspace

The agent lists your brand kits, caption templates and connected destinations before it touches a video.

Discovery

sendshort workspace:list

# Inspect a brand kit

sendshort brandkits:get bk_7Ka2

# Available caption templates

sendshort templates:list

Create the clips

One call imports the source, finds the strongest moments, burns captions and renders every ratio you asked for.

Clip creation

sendshort clips:create \

--source "https://youtu.be/xxxxxxxx" \

--clips 5 \

--ratio 9:16,1:1,16:9 \

--template "brand-kit" \

--language en

Review, then publish

Clips land in your SendShort workspace. Download, tweak in the editor, or hand them to your scheduling stack.

Delivery

sendshort clips:list prj_9fA2

# Download the finished renders

sendshort clips:download prj_9fA2 --all

# Or forward to your scheduler

sendshort clips:export prj_9fA2 --to postiz

From any chat

Ask in plain language, get clips back

No timeline, no exports, no file juggling. Describe the shorts you want and the agent runs the whole pipeline — import, clip, caption, reframe, render.

ChatGPT

New chat

Images

Library

Plugins

Projects

More

youtube.com/watch?v=…

Import this YouTube video and make 5 shorts.

Done - prj_9fA2. Imported the video and rendered 3 high-retention clips:

SendShort agent: Done — prj_9fA2

00:42 "Hook" 9:16 · captions ✓

03:18 "Story" 9:16 · captions ✓

07:05 "Lesson" 9:16 · captions ✓

2 more rendering, webhook on finish.

Captions burned in · ready to download or keep editing in SendShort.

Creator speaking into a microphone, captioned short

I shared this story

0:22

03:18 — Story

Creator gesturing to camera, captioned short

go to my website

0:48

14:50 — CTA

Creator in shirt and tie speaking, captioned short

proof that it works

0:31

11:22 — Proof

Ask for a different aspect ratio, language, or caption style…

Works with the assistants you already use

Claude

Desktop, Code and the API

Grok

Clip inside your editor

Codex

Shell-driven automation

ChatGPT

Connector for Plus and Team

Hermes

Chat-app driven workflows

OpenClaw

Runs local, posts on a schedule

Raw → post-ready

Why AI agents work well with SendShort

Same footage, reframed to 9:16, cut to the moment that holds attention, captioned word by word and styled to your brand kit — no timeline, no export settings.

Every tool call returns structured JSON your agent can parse

A bundled SKILL.md documents every command and argument

Discovery lets the agent read brand kits before rendering

Long jobs are async with webhooks — no timeouts mid-render

Model agnostic: Claude, GPT, Gemini or open-source LLMs

Human in the loop by default — nothing publishes without you

SKILL.md (auto-discovered)

---

name: sendshort

description: Turn long videos into

captioned shorts from chat

metadata:

requirements:

env:

- SENDSHORT_API_KEY

---

# Available tools

- sendshort workspace:list

- sendshort templates:list

- sendshort clips:create

- sendshort clips:list <project>

- sendshort clips:download <project>

Tool reference

The same surface over MCP or the CLI. Every call maps to a REST endpoint on the API.

Tool

Description

workspace:list

List workspaces, brand kits and destinations

templates:list

Fetch available caption templates

clips:create

Import a source and render shorts

clips:list <project>

List clips and their render status

clips:download <project>

Download finished renders

clips:export <project>

Send clips to a connected scheduler

translate <clip>

Dub and caption into 100+ languages

Give your assistant an editing team

API and AI Agent are included on Business - everything else starts free

See pricing