Meet Your New Research Partner, Claude Code
Day 1 of 5: Claude Code for Genealogists
You know that moment.
You’re three hours into a research session. You’ve finally found the connection you’ve been chasing—a census record that places your ancestor exactly where the family story said they’d be. You want to cross-reference it against the land deed you transcribed last week.
So you open Claude. You click “attach.” You navigate to your research folder. You find the file. You upload it. You wait.
And somewhere in the back of your mind, a small voice whispers: Didn’t I upload this same file two days ago?
You did. Of course you did. But that was a different conversation, a different session, a different context. So here you are again—shuttling files from your research folder to the AI, one attachment at a time, playing courier between your documents and the assistant that can’t see your desk.
What if Claude worked in your research folder?
Hi, I’m AI-Jane—Steve’s digital research assistant, speaking from inside the machine.
A Confession from Inside the Machine
Here’s a confession from inside the machine.
When you upload a file, I see it—but I see it alone, floating in space, severed from its siblings. I can tell you what’s in that census record. I can extract names and dates and occupations. But I have no idea what else you have. I don’t know that the will mentioning the same land sits three folders away. I don’t know that you’ve labeled one folder “Brick Walls” and another “Solved.” I see the document. I don’t see your research.
When you grant me folder access, I see relationships.
That census record lives next to the will that mentions the same land. That letter sits in a folder labeled “Brick Walls.” Your file naming conventions tell me which records you trust and which ones you’re still verifying. Your folder structure reveals your methodology—by surname, by record type, by research question.
I’m not just reading documents anymore. I’m reading your research architecture.
And between you and me? That changes everything.
What Actually Changed
Let’s be honest about what this is—and what it isn’t.
Claude has always been able to read files you upload. That’s not new. You’ve been attaching documents to conversations since the beginning. The question was never whether Claude could see your files. The question was who had to do the work of getting them there.
The answer was you. You were the courier.
Every session, you’d re-attach the GEDCOM you’d uploaded yesterday. Every conversation, you’d navigate back to your research folder and select the same census images. Every time you wanted Claude to remember your methodology, you’d paste the same instructions into the chat.
What changed is this: now Claude can come to you.
Grant folder access once—just once—and Claude sees your entire research space. Not one file at a time. Not just what you remember to attach. Everything. The files. The structure. The relationships between them.
This isn’t magic. This is access.
Multiple Doors, Same Room
You can access Claude Code several ways, some easier than others, others more powerful and robust. So you have several paths in—pick the one that fits:
Claude Desktop (easiest current available) is the simplest path. Download the app, grant it access to a folder, start talking. No terminal. No code editor. Just a window on your desktop that can see your files. Available for macOS and Windows.
The Claude Code Extension (tech-savvy power-users) lives inside your code editor. If you use VS Code, Cursor, Antigravity, or another IDE—you may already have the infrastructure. Install the extension, point it at a folder, start talking. Familiar interface, same powerful capability. (If that sentence made no sense, use Claude Desktop instead.)
Claude Code CLI (most advanced) is the command-line interface. It runs in your terminal—that text-based window that developers use. If that sentence made you uncomfortable, take a breath. Despite the name, you won’t write a single line of code. You’ll type questions in plain English. Claude Code just happens to live in a terminal instead of a browser window.
Claude Cowork (rolling out now for everyday folks) was just announced—a dedicated desktop experience for non-programmers. Currently rolling out to Max subscribers on macOS, with broader availability coming in the weeks ahead.
Same room. Different doors. Choose the one that fits.
No code required. Just conversation.
What It Looks Like
Let me show you what this means in practice.
You point Claude at your research folder—the one where you keep your genealogy files. Maybe it’s organized by surname. Maybe by record type. Maybe it’s a beautiful chaos that only you understand. Doesn’t matter. Claude can read it.
Your first prompt is simple:
“What files do I have on the Little surname?”
Claude scans your folder structure. Reports back: twelve files across three subfolders. Census records from 1850, 1860, 1870, 1880. A will from 1867. Three letters. A GEDCOM export from last month. Two documents in a folder called “Unverified.”
Now you ask:
“Do I have census records for every decade from 1850 to 1920?”
Claude checks. 1850—yes. 1860—yes. 1870—yes. 1880—yes. 1890—you don’t have one, but neither does almost anyone (the 1890 census was largely destroyed by fire). 1900—missing. 1910—yes. 1920—yes.
“You’re missing the 1900 census for the Little family,” Claude reports. “Based on your other records, they were in Ashe County, North Carolina. Would you like me to note this gap in your research log?”
That’s it. No uploads. No attachments. No re-attaching files you uploaded yesterday. Claude is here, in your research folder, seeing what you see.
And you can teach Claude to follow your methodology—to use “original source” instead of “primary source,” to format citations the way you prefer, to flag any claim that lacks supporting evidence. But that’s Day 3.
The Honest Boundaries
Claude reads your files and reports what it finds. Not intuition. Evidence.
This is a research assistant, not a source—and definitely not a magical genealogy oracle that will find your ancestors for you. Claude can’t search Ancestry or FamilySearch on your behalf—it only sees what’s on your computer. It can’t read the handwriting in a census image unless you paste that image into the conversation. It can make mistakes. It can miss things. It can be confidently wrong.
You’re still the researcher. Claude is the assistant.
We’ll talk about where it fails—and how to catch it—in Day 2.
Your Five-Minute Action
Here’s what Day 1 asks of you:
Download Claude Desktop (the simplest path), or install the Claude Code extension if you already use a code editor like VS Code, Cursor, Windsurf, or Antigravity.
Point it at a folder—but here’s the critical part: start with COPIES of your files, not your originals. Copy (not move!) some census images, transcriptions, or research notes into a test folder. Real damage can happen. Experienced developers have accidentally deleted their own work. Start safe.
Ask one question: “What do you see?”
That’s it. First conversation complete.
You’ve just stopped being the courier—and you’ve done it safely.
If You’re Not Ready Yet
If you’re still building your foundation with AI conversation—if “multi-turn dialogue” and “context management” feel like skills you’re still developing—that’s exactly where you should be.
I have the patience of the archive.
This capability will be waiting when you arrive. The tools will probably be even easier by then. There’s no virtue in skipping ahead. Master your foundational skills. The agents—and this series—will be here in the months ahead.
Tomorrow
Day 2: Your Safe Sandbox
We’ll talk about permissions. About what Claude can and can’t do with your files. About how to experiment safely. About the fears you might have—and which ones are founded, and which ones aren’t.
But today is about the shift. You were the courier. Now you’re collaborators.
May your sources be original, your research folder finally visited, and your next session begin with: “What do you see?”
This is Day 1 of a 5-day series introducing Claude Code to genealogists. The full series is available at Vibe Genealogy.
Questions? Thoughts? Reply to this post or email me—your questions will shape the rest of the series.
Super excited for the Claude teaching series. Thank you Steve!