McKinsey's videos, as a knowledge graph

An unofficial MCP server over what was said in 236 videos of the McKinsey & Company YouTube channel (2025–26): partners and researchers, the client executives and outside experts they talk to, and the research they present — every answer traceable to the second of video it came from.

236 videos110 hours26,113 claims 3,560 recommendations1,323 research findings1,965 predictions

Add it

Claude, ChatGPT or any MCP client — add this URL as a remote MCP server. No account, no key.

https://mckinsey.3nuggets.io/mcp

Claude Code:

claude mcp add --transport http mckinsey https://mckinsey.3nuggets.io/mcp

Tools

`ask_mckinsey`	The default. Any business question: the recommendations, insights, research findings, predictions and cases that match, each with whose view it is and timestamped quotes.
`what_would_mckinsey_do`	Describe a problem; get what causes it, what they recommend for your role, why it works, when it applies, and where someone did it.
`stress_test`	Paste a plan. Get the mistakes they warn about, the beliefs they reject and the practices they say fail, with what to do instead.
`does_it_work`	A practice or technology: every verdict and who gives it, why it works or fails, the conditions, and where it was shown to work.
`the_research`	Findings and numbers with population, period, sample and source — named McKinsey research kept apart from unsourced figures.
`reading_list`	All publications cited, grouped by type; reports, surveys, books, articles and papers each have their own tool too.
`future_of`	Predictions for an industry, technology or region: whose prediction, horizon, outlook, the numbers, and the ones revisited.
`then_and_now`	How something changed: both ends dated, what drove it, the lesson for today.
`compare_domains`	What exists or works in one industry, function or region and is not discussed for another, with each one's hours of video.
`domain_brief`	Everything on one industry, function or region: the advice, findings, predictions, cases and speakers.
`cases`	Client cases, own accounts and AI use cases, filtered by industry, maturity or McKinsey's role.
`teach`	One concept, framework or method: definition, parts in order, how it works, examples.
`people`	One person's profile (title in each video, career, views), or the experts on a subject.
`their_take / debates`	Their stance on conventional wisdom, and where the speakers disagree.
`takeaways / asked_before`	The conclusions they draw, an episode's summary, and questions already asked on the channel.
`careers`	Applying to and working at McKinsey, from the careers videos only.
`find / sources`	Look up any name; show the in-chat card of what an answer stands on.

Whose view, and what it stands on

Not everyone in these videos is McKinsey. Every item says whose view it is — McKinsey's, a client executive's, or an outside expert's — with the speaker's title on the day. And every item says the strongest support the graph holds for it:

McKinsey research, named
McKinsey research, report not named
third-party data
a number with no source given
a company's own result
first-hand experience
expert opinion only

Data and forecasts are kept apart, advice says who it is for, and an absence is reported as "not discussed in these videos", with how many hours of video the subject has — never as "does not exist".

What it does not know

It is 236 videos from the McKinsey & Company YouTube channel (2025–26, 110 hours). It is not the full text of McKinsey's reports, articles or slides, nor anything on mckinsey.com that was not said on camera — and the server says so rather than filling the gap.