VideoCore
← Back to VideoCore
MODEL CONTEXT PROTOCOL · NEVER TOUCHES YOUR CHANNEL

VideoCore, inside your AI assistant

Connect VideoCore to Claude or ChatGPT and ask in plain words: "the five biggest channels on my topic, and which of their videos beat their usual numbers", "why isn't my latest video getting views?", "audit this channel and find its breakout videos", "is this niche worth it?". The answers start from public YouTube numbers; the comparisons against each channel's own uploads are calculated by VideoCore and every answer says which figures are ours, not YouTube metrics. Free on every plan: 30 lookups a day on Free, more on paid plans, with no credits to buy and an allowance that resets every day, and every answer links to the full view in VideoCore.

Create a free account Try the free video diagnosis

Paste one of these

Prompts that read your own channel

Connect VideoCore in Claude or ChatGPT (steps below), press Copy, paste it into a new chat and fill in the part in brackets. Each one calls VideoCore's tools on your channel, so the answer is about your videos, not general advice.

Why is my last video flat?

A plain answer on your latest upload, set against your own channel's usual numbers, and the first thing to change.

Using the VideoCore connector, diagnose my latest video: [paste the link of your latest video]. Tell me in plain words why it is or is not getting views compared with my own channel's usual numbers, and the one change to the title or thumbnail I should make first. Quote only figures VideoCore returns, and say which ones VideoCore calculated.
Uses diagnose_video
Which of my videos should I remake?

Your videos that beat your usual numbers and the ones well below them, with two picked to remake.

Using the VideoCore connector, find the outlier videos on my channel [your channel link or @handle]: the ones that beat my own usual numbers and the ones well below them. Then pick two of the ones below to remake with a new title and thumbnail, and tell me why for each, using only what VideoCore returns.
Uses find_outliers
What should I make next?

An audit of your channel, what is moving in your niche now, and three ideas with working titles.

Using the VideoCore connector, audit my channel [your channel link or @handle], then check what is moving in my niche right now and give me three next video ideas, each with a working title and why it fits my channel. Say which facts came from YouTube and which VideoCore calculated.
Uses channel_audit, trending_in_niche, video_ideas
Better titles for my next video

Title patterns from the real top videos on your topic, turned into titles in your own style.

Using the VideoCore connector, get title ideas for my next video about [your topic], built from the real top-performing videos on that topic. Write five titles in the style of my channel [your channel link or @handle], and show me the real videos each pattern came from.
Uses keyword_ideas

Working on your next video

Start from your own best videos, check a title before you publish, or talk your idea through and get a draft script in your own words.

What does my viewer really want?

From what your best videos teach to what the viewer wants for their life, and three titles that promise that.

Using the VideoCore connector, find the outlier videos on my channel [your channel link or @handle]. For each one that beat my own usual numbers, keep asking "why does the viewer want that?": from what the video teaches, up to what the viewer wants for their life. Stop at the step a viewer would say out loud in their own words, not the most abstract one. Tell me what that viewer is turning over in their head late at night when they search for this, then write three titles for my next video that promise that change, not just the information. Use only what VideoCore returns about my videos.
Uses find_outliers
Would my title turn heads?

Your title checked by VideoCore, then bolder versions built from the real top videos on the topic, each one your video can deliver.

Using the VideoCore connector, run check_title on my title: [your title]. Then get keyword ideas for its topic. Tell me plainly whether my title sounds balanced and reasonable, the kind that turns no heads, and write five bolder versions that take a clear side or set up a contrast, using the patterns in the real top videos VideoCore returns. Only promise what my video actually delivers. Run check_title on your two best versions and show me the results.
Uses check_title, keyword_ideas
Talk it out, get a script in your own words

Your assistant interviews you about your next video one question at a time, then drafts the script from what you said.

Using the VideoCore connector, list my projects and ask me which one we are working on (or the topic, if it is not there yet). Then interview me about it, one question at a time, waiting for each answer: who it is for, what they get, the story behind it, the mistake people make, and what I would tell a friend about it. When I say "done", draft the script in my own words and phrases from my answers wherever you can, mark anything you added, and add no facts or numbers I did not give you.
Uses my_projects

Routines to schedule

If your assistant can run a prompt on a schedule, set one of these to run at the time in its title; if not, paste it yourself that morning. Each run uses some of your daily lookups, and every answer says how many are left.

Every Monday: last week against my usual

What beat your own usual numbers last week, what fell below, and one thing to do this week.

Using the VideoCore connector, read my channel [your channel link or @handle] with channel_audit and find_outliers. List my videos published in the past week that beat my own usual numbers and the ones well below them. Leave out videos too new to judge or too close to usual, and say which you left out. For each winner, say what it shares with my other winners; for each one below, suggest a new opening line. End with one thing to do this week. Use only figures VideoCore returns, say which ones VideoCore calculated, and give me the open_in_videocore links.
Uses channel_audit, find_outliers
Every Thursday: what is moving in my niche

What is picking up in your niche this week, which of it fits your channel, and two ideas.

Using the VideoCore connector, check what is trending in [your niche], then get video ideas for it. Tell me which of the fast-moving videos fit my channel [your channel link or @handle] and which do not, and give me two ideas with working titles. Say which facts came from YouTube and which VideoCore calculated. If lookups_left_today runs low, stop and tell me.
Uses trending_in_niche, video_ideas
First of the month: my seasons

Whether the month ahead has been above or below usual on your own channel, and whether your seasonal titles help.

Using the VideoCore connector, read my channel's seasons with my_seasons: the months first, then the titles. Tell me whether the month ahead has been above, below or usual for my channel in the years VideoCore measured, and whether my titles with a season or a year in them did better or worse than my others. If VideoCore says there is not enough history yet, say so and stop. This is my own history, not a forecast: do not promise views.
Uses my_seasons

What you can ask

Tools built on public YouTube numbers, one connection

diagnose_video

Why is this video under/over-performing — YouTube's counts, compared by VideoCore with the channel's own median, + an AI thumbnail teardown.

channel_audit

Subscribers, upload cadence, recent median views, consistency, best & worst recent videos — for any channel.

find_outliers

A channel's breakout videos (views ≫ its own median) — the topics/formats actually working right now.

find_niche

Score a niche from its top videos: demand, competition, a rough RPM estimate and trend, with a viability score calculated by VideoCore.

trending_in_niche

Recent uploads in a niche ranked by view velocity — momentum before it peaks.

video_ideas

Specific next-video ideas grounded in what's overperforming in your niche right now.

keyword_ideas

Title angles built from the patterns of the real top performers for a topic.

compare_channels

Benchmark 2-4 channels side by side on size-adjusted figures calculated by VideoCore.

check_title

VideoCore's measured title checks: who plus what, the thumbnail's words against the title, and your opening against it. Never a score.

my_seasons

Your own channel month by month against its usual, by year, and how your titles with a season or a year in them did.

my_projects

The videos you are working on in VideoCore, so your assistant knows what you mean by "my next video".

Setup

Connected in under a minute, free on every plan

Add the connector

In Claude: Settings, Connectors, Add custom connector. Name it VideoCore and paste this address:
https://videocore.uk/api/mcp.php

Press Connect

VideoCore opens. Sign in, pick your workspace and press Connect. There is no key to copy.

Using Cursor, VS Code or Claude Code?

Those take a key instead. Make one under Settings, API Keys and add it as a Bearer header.

Ask away

"The five biggest channels on budget van builds, and which of their videos beat their usual numbers", "diagnose youtube.com/watch?v=… ", "audit @mkbhd", "what's trending in home espresso?", "give me 5 video ideas for budget van builds."

Free Claude skills

Teach your assistant to use VideoCore

videocore-creator: which VideoCore tool to use for which question about your own channel. Ask "is this title good?" and it runs the title checks; "when should I publish?" and it reads your own seasons; "why is my video flat?", "what is working on my channel?" or "what should I make next?" and it picks the right lookups in order. It never makes up a number, never turns a check into a score, and says plainly that your own history is not a forecast.

It needs the VideoCore connector connected (above). Each lookup it makes counts towards your daily lookups, as it would if you asked yourself. Install it the same way as the weekly review below; it works in Claude and in other assistants that read skill files.

Download the skill (.zip)

A weekly review in one request

A skill is a set of written steps Claude follows. This one, videocore-weekly-review, runs four VideoCore lookups in order: an audit of your own channel (by its @handle), your breakout videos, what is moving in your niche, and next-video ideas. Then it writes a one-page plan: the facts first, then what to do this week. It never makes up a number: every figure comes from a VideoCore answer, and it gives you the link to each full view in VideoCore.

It needs the VideoCore connector connected (above). One review uses about 7 of the 30 lookups a day on the Free plan, about 8 if you add a channel to compare with.

Install it in Claude: first make sure "Code execution and file creation" is switched on in Claude's settings. Download the zip, then in Claude go to Customize, Skills, press +, then Create skill, then Upload a skill, and choose the zip as it is (don't unzip it). Then ask: "run my VideoCore weekly review". In Claude Code, unzip it into your skills folder (~/.claude/skills/).

Download the skill (.zip)

Safe by design. The connector reads public YouTube data and your own VideoCore projects. It can never post, change settings, or touch your channel. Each key is yours alone and revocable anytime from Settings.

Powered by VideoCore · Video Diagnosis  |  Niche Finder