Does it upload to YouTube for me?

No. The finished file stays on your disk and its title and description are saved as a draft for you to review. You upload. Publishing isn't part of the server today, which also means nothing reaches your channel without you seeing it first.

Does it do voiceover?

Not built in. Films come with a synthesised soundtrack mastered to about −14 LUFS, and your own voice track drops onto the timeline like any other audio. If you use a text-to-speech service, the agent can place its output; the voice itself comes from you or that service.

Is this a Remotion MCP?

No, though the idea is close: video as code, rendered frame by frame through a headless browser into ffmpeg, with a live preview you can scrub before the render. NotPeople's recipes use their own React/canvas scenes, Playwright captures and Blender for 3D.

Which MCP clients work?

Claude Code, Codex and Grok Build have one-line install commands on this page. Claude Desktop, Cursor and any MCP client that speaks HTTP transport work the same way.

One plan, $99/mo: the research tools, drafts, every playbook including clips, 300 Google and YouTube pages and 300 GPU render minutes a month. AI stills and AI video clips run on your own keys (Gemini, Picsart), and your own vast.ai key lifts the render cap, so you pay those providers directly. The first 3 days are free, no card. Cancel anytime.

YouTube MCP server · Claude Code · Codex · Grok Build

YouTube MCP
that makes the video.

A YouTube MCP server for Claude Code, Codex or Grok Build. Your agent finds what's overperforming in your niche, reads the transcripts, cuts your long videos into captioned Shorts, draws thumbnails and renders films from code. The heavy work runs on your own machine, and nothing reaches your channel until you upload it.

$ claude mcp add --transport http notpeople https://mcp.notpeople.ai/mcp --header "Authorization: Bearer $NOTPEOPLE_KEY"

Get your key — free Watch it work

Research · Shorts · films·Sources stay on your disk·X & Reddit side →

claude — shortsffmpeg · local

❯ cut last week's 47-min episode into shorts. captions big, my usual style

⏺ aicb_get_playbook(kind: video)

⏺ aicb_memory_search(workflow:video · style:clip)

house style: punch captions · 35–55s · no intro card

⏺ yt-dlp(youtu.be/… → episode_047.mp4 · 47:12)

⏺ faster-whisper(local · 2–4s segments)

moments that land without the setup:
41  the part everyone clipped was not the part that mattered
18  why we killed the waitlist
02  numbers, no adjectives        · +3 more

● Plan: 5 shorts, 9:16 over a blurred fill, punch captions. Go?

❯ go

⏺ ffmpeg(cut · reframe 1080×1920 · burn · one at a time)

5 files · frame checked mid-clip · captions rendered

⏺ aicb_save_draft(5 × title + description · pending)

drafts ready · files in ~/Videos/ep047_shorts · you upload

Renders on your machine

Your footage never leaves your disk.

Your agent directs, the NotPeople MCP server holds the recipes, briefs and drafts, and the heavy work runs on your machine: yt-dlp, a local speech model, ffmpeg, the timeline editor and the renderer. Sources and renders stay on your disk; only the finished file goes up for review.

Your agentyour laptop

Claude Code · Codex · Grok Build

cut this 47-min episode into 5 shorts

reading the transcript · 6 moments…

NotPeople MCPnotpeople cloud

playbooksvideo · animation · gen

series3 formats

memorycaption style · length

drafts5 shorts pending

Local renderyour machine

ffmpeg · timeline · 1080×1920

V2

V1

A1

short 3 / 5 · 41srendering 68%

source → windows → cut → captions → render → draft

Local render

your CPU or GPU, your files

Plan before pixels

the cut is agreed in words first

House style, remembered

caption look, length, format per series

What it does for YouTube

A production team, as tools your agent can call.

Find what's working

aicb_youtube_search · aicb_youtube_shorts · aicb_youtube_video

niche: ai workflows · last 30 days · long + shorts3 outliers

views ÷ channel mediansmall channels, big numbers

I let an AI agent run my channel for 30 days14k subs29×

the MCP setup nobody talks about3.1k subs18×

Claude Code made my B-roll (no stock)41k subs6×

10 AI tools for YouTuberscrowded1.1×

transcripts read · 3 hook patterns foundbrief drafted

Long video → Shorts

yt-dlp · local whisper · ffmpeg

THIS WAS THE PART THAT MATTERED

youtu.be/… · 47 min · transcript6 moments found

12:41the part everyone clipped was not the part that mattered

04:18why we killed the waitlist

27:02numbers, no adjectives

33:50the question nobody asked

cut 12:41 → 13:22 · 41s · 1080×1920subs burned · queued 09:40

Thumbnails

aicb_gen_cover

30 DAYS

NO STOCK

29×

3 concepts · 16:9pick one

Captions, any language

local transcript · one cut, many burns

EN · burnedES · burnedPT · burnedDE · nextJA · next

ESTA ERA LA PARTE QUE IMPORTABA

same timing, translated text

Films drawn in code

animation · terminal · site_film

animation · product film · data filmscore synthesised · −14 LUFS

16:91920×1080 · 60 fpsup to 60 s

9:16native vertical≤ 30 s

1:11080×1080feed

GIF2–6 s loopno sound

A real timeline

OpenChatCut over MCP · local

V2

V1

A1

remove silencecut filler wordslandscape → portraitcaptionsnormalize loudness

you watch in your window · the agent editsexport pending

Also:long renders on a rented GPU the agent starts and stops·AI b-roll and stills on your own key·public-domain archive footage & foley·the upload stays yours

Watch it work

Three things you'll ask in the first hour.

claude — shortsffmpeg · local

❯ cut last week's 47-min episode into shorts. captions big, my usual style

⏺ aicb_get_playbook(kind: video)

⏺ aicb_memory_search(workflow:video · style:clip)

house style: punch captions · 35–55s · no intro card

⏺ yt-dlp(youtu.be/… → episode_047.mp4 · 47:12)

⏺ faster-whisper(local · 2–4s segments)

moments that land without the setup:
41  the part everyone clipped was not the part that mattered
18  why we killed the waitlist
02  numbers, no adjectives        · +3 more

● Plan: 5 shorts, 9:16 over a blurred fill, punch captions. Go?

❯ go

⏺ ffmpeg(cut · reframe 1080×1920 · burn · one at a time)

5 files · frame checked mid-clip · captions rendered

⏺ aicb_save_draft(5 × title + description · pending)

drafts ready · files in ~/Videos/ep047_shorts · you upload

Downloaded, transcribed and cut on your machine. The plan is agreed in words before anything renders; the files and their titles land as drafts for you to upload.

Playbooks

Every job has a recipe. Your agent reads it, then works.

Every multi-step job is a chain of tool calls. The playbook is the order and the fields — so nothing gets guessed. Override any recipe for your account with one memory note.

1

aicb_get_playbook

kind: video

2

aicb_youtube_video

transcript · 6 moments

3

yt-dlp + whisper

local · 2–4s segments

4

ffmpeg

cut · 9:16 · captions

5

aicb_gen_cover

thumbnail · 3 concepts

6

aicb_save_draft

title + description

Pricing

3 days free. Then one plan.

Trial

$0 for 3 days

Everything in Pro, before you pay.

✓The full Pro plan for 3 days
✓One per email, no card
✓Starts on the Connect page after Google sign-in
✓Subscribe in the workspace to keep going

Start the trial

POPULAR

Pro

$99 / month

One plan, everything in it. Cancel any time.

✓Unlimited accounts, and the people who run them with you
✓The content timeline: drafts, approvals, series, memory, media
✓Monitorings for mentions, keywords and comments, checked every 15 min
✓82 MCP tools for your Claude
✓1,500 cloud reads a day: X, Reddit, Threads, Instagram, video links
✓300 Google and YouTube pages a month, unmetered on your own SerpApi key
✓A residential proxy exit on 5 accounts
✓50 GB of media storage
✓300 GPU render minutes a month, or your own vast.ai key

Get Pro

LiveX · Reddit·Read-only Threads · YouTube · Instagram · Polymarket·Requires Claude Code, Codex or Grok Build · macOS / Linux / Windows

Card by Stripe · Discount codes on the payment page · Cancel any time · No card for the trial

Get your key

Sign in. One line to install.

The key is yours the minute you sign in: it identifies you, so your agent sees your accounts and nothing else. Setup takes about 15 minutes from there: the step-by-step guide.

$ claude mcp add --transport http notpeople https://mcp.notpeople.ai/mcp --header "Authorization: Bearer $NOTPEOPLE_KEY"

1
Sign in with Google
Open app.aicb.notpeople.ai. No password, no form — your Google account is the login.
2
Start the 3-day trial or subscribe
The trial is on the Connect page, one per email, no card. Pro is billed in the workspace.
3
Generate your key
Connect → Generate my key. It comes with the install line below and the first message to paste into your agent.

Open the workspace3-day trial · no card · cancel anytime

FAQ

Questions, answered.

An MCP (Model Context Protocol) server gives an AI agent tools it can call. A YouTube MCP server lets Claude Code, Codex, Grok Build or any MCP client work with YouTube: search videos and Shorts, read a video's details and transcript and, in NotPeople's case, run the production work around it (cutting, captions, thumbnails, generated films) on your own machine. You add it with one line and ask in plain language.

Most YouTube MCP servers do one of two things: read data (search, stats, comments, transcripts through the YouTube Data API or yt-dlp) or publish through the API. NotPeople does the research part too, and then makes the video: it finds the moments in a long episode, cuts and reframes them into Shorts with burned captions, translates the captions, draws thumbnails, and renders animated, product and data films from code. Each job follows a written playbook, so the agent works in a fixed order and doesn't improvise.

Shorts cut from your long videos or podcasts, 9:16 with captions in your style and in several languages; thumbnails; code-drawn animation from a 3-second loop to about a minute, at 1920×1080 60 fps plus native vertical, square and GIF versions; 30–50 second product films shot from real captures of a website; and short data-driven films built from real market or system data. Longer edits of your own footage go through a local timeline editor the agent drives over MCP. Generated films top out around a minute, and masters are 1080p.

ffmpeg and yt-dlp, which the setup playbook installs. Transcription runs on a local speech model. The timeline editor needs Node 24 and is comfortable from 16 GB of RAM; the agent asks before installing it. Animation and product films render on a CPU. 3D scenes and the fastest data-film path want an NVIDIA GPU. Without one, the agent can rent a GPU box on Vast.ai with your key, run the render and shut the box down; a ten-minute test render cost us $0.06.

Your move

One line. A production team.

3 days free, no card. Then $99/mo for research, clips and every playbook. Cancel anytime.

Get your key X & Reddit MCP server

YouTube MCPthat makes the video.