Someone who watched my last video left a comment I get a lot: "But how do I add YouTube and Gemini API as connectors?" It's a fair question, and the honest answer starts with a small correction, because the way connectors work in Claude isn't quite the way most people picture it.
Claude has no ready-made YouTube or Gemini connector to switch on. You add them yourself, one of two ways: as a no-code custom connector (Settings, then Connectors, then Add custom connector, pointing at a hosted MCP server), or inside Claude Code with claude mcp add and your own API key. Gemini is worth flagging early, because it's a model API rather than a data source, so it doesn't behave like a connector at all.
What a connector actually is
A connector lets Claude reach an app or service, read your data there, and take actions on your behalf. Slack, Google Drive, Linear, Notion: those are connectors, and they all run on the same plumbing underneath, called MCP (the Model Context Protocol). You'll find the ready-made ones in the Connectors Directory, each with its own page describing what it can read and write.
The thing to understand is that a connector wraps a service in a small server Claude knows how to talk to. When there's an official one, you click Connect and sign in. When there isn't, which is the case for both things the commenter asked about, you point Claude at a server yourself. That's the whole game, and once it clicks, "how do I add X" stops being scary.
The YouTube API: two honest routes
There's no official YouTube connector, but YouTube data is a genuine connector job: you want Claude reading video stats, comments, and search results. So you bring your own server. Which route you pick depends on whether you'd rather never see a terminal or you're happy in one.
Route 1 — No-code, in the Claude app
If you live in the Claude desktop or web app, you can add YouTube without writing anything. Go to Customize, then Connectors, click the plus, and choose "Add custom connector." You give it a name and the URL of a hosted YouTube MCP server (several exist, run by other people, so vet the one you choose the way you'd vet any tool that touches your data). Claude connects to that server from Anthropic's cloud, so the server has to be reachable on the public internet, and on the free plan you get one custom connector, which is worth knowing before you spend it.
This is the route I'd point most non-technical people at first. It's the closest thing to the experience you were probably imagining when you asked the question.
Route 2 — Claude Code, with your own key
The route I actually use for agency work is Claude Code, the version that runs in your terminal and wires into everything else. Adding a server is one line:
That registers a local server, passes your API key in as an environment variable, and makes the tools available next session. Run claude mcp list to check it says connected. If you'd rather point at a hosted server over the web instead of running one locally, the shape is claude mcp add --transport http youtube <url>, and a token goes in with --header rather than -e.
The key itself comes from Google. In the Google Cloud Console you create a project, enable the YouTube Data API v3, then generate an API key under Credentials. That key is what the -e YOUTUBE_API_KEY= part is feeding the server. One setup note that saves confusion: servers are added at a scope. The default is local, meaning it's active only in the project folder you ran the command in. Add --scope user and it follows you into every project, which is usually what you want for something like YouTube.
The Gemini API: not really a connector
Here's the correction I promised. Gemini isn't a service you connect to the way you connect YouTube; it's a rival model's API. Connectors exist to give Claude data and actions, and a second model isn't data. So "add Gemini as a connector" doesn't map onto anything, and for most people the useful answer is that you don't need it. Claude is doing the orchestration in these workflows, and the connectors that earn their place are the ones bringing in your own data.
There's one real exception, and it's why the API came up in my video at all. Claude reads text; it can't watch a video frame by frame. Gemini can, including the words shown on screen. So in some of my research workflows I do call the Gemini API for exactly that one job, reading what a video shows rather than only what it says. If you want that, you wire the Gemini API in as a tool through an MCP server or a small script, with the same environment-variable pattern as above. The key comes from Google AI Studio: sign in, click "Get API key," create one, and copy it. There's a free tier to start on.
So the two things the commenter bundled together are actually different jobs. YouTube is a connector you supply yourself. Gemini is a specialist tool you occasionally call for a task Claude can't do alone.
Keep your keys out of trouble
One rule matters more than the rest: treat an API key like a password. Anyone who has it can spend your quota and run up your bill. Never paste a key into a public chat, a screenshot, or a shared doc, and never commit one to a Git repository. The reason connectors take keys through environment variables and header flags, rather than sitting in your prompt, is precisely so the key stays out of places it can leak. Follow that grain and you'll be fine.
If you want the wider context these fit into, I've written up the five Claude workflows I actually run, the SEO system built on these connectors, and a broader look at how a small business can automate real work with them.
If you'd rather have someone set this up around your team than wrestle with it yourself, that's part of what we do at Humble&Brag. Send me the workflow you're trying to build and I'll tell you honestly whether a connector is the right tool for it.
FAQ
Does Claude have a YouTube connector? Not an official one. You add YouTube yourself, either as a custom connector in the Claude app pointing at a hosted MCP server, or in Claude Code with claude mcp add and a YouTube Data API v3 key from the Google Cloud Console.
How do I add a custom connector in Claude? Go to Customize, then Connectors, click the plus, and choose "Add custom connector." Give it a name and the URL of the MCP server you're connecting to. On the free plan you can add one custom connector.
Can I connect the Gemini API to Claude? You can call it, but it isn't a connector in the normal sense, because Gemini is a model rather than a data source. Most workflows don't need it. The exception is reading what a video shows on screen, which Claude can't do, so I sometimes call Gemini for that one task.
Where do I get a Gemini API key? From Google AI Studio. Sign in with a Google account, click "Get API key," create one, and copy it. There's a free tier to begin with.
What's the difference between a connector and Claude Code? A connector is the thing you're plugging in; Claude Code is one place you plug it in. The Claude app takes connectors through its settings; Claude Code takes them through the claude mcp add command. Both use MCP underneath.
