Giving Claude personalisation superpowers with my Google search, YouTube, and Instagram history
Using Fabric MCP in Claude to connect my rich, personal context
I plugged my Google Search, YouTube, and Instagram history into Claude using Fabric MCP and now it can plan my upcoming trip proactively, recommend movies and books I will actually consume, and even chat with me about my Instagram stories like a friend.
TL;DR
Connect Fabric MCP to Claude with authentication in a few clicks so Claude can use your real context from Google Search, YouTube history, Instagram posts and stories.
The responses start to reflect you. Claude spotted my upcoming Capri trip, built an itinerary I will use, identified a rare bird I rescued on Sunday morning, and tied it to my late night wine bar visit.
This is not a Claude only thing. Fabric MCP works anywhere MCP works, including ChatGPT and Cursor.
What Fabric MCP is and how it connects to Claude
Fabric MCP is the bridge between your data sources: Google, YouTube, Instagram, and Gemini and MCP-compatible apps like Claude and ChatGPT.
You stay in control. You choose what to connect and you grant permission. Once connected, the model starts reflecting your preferences without long, repetitive prompts.
Fabric MCP does the heavy lifting that makes your data usable by the model. It normalises messy exports, adds asset descriptions for images and videos, attaches timestamps and source links, enriches interactions, and exposes clean tools the agent can call.
Connecting takes minutes:
In Claude, open Settings → Connectors, then click Add custom connector.
Paste the Fabric MCP details.
Click Add and authenticate.
What changes after connecting Fabric
Test 1: Recommending books and movies
I ran two Claude chats side by side. One had Fabric MCP. One did not.
Without context: Claude defaulted to safe, generic ideas.
With context: Claude sounded like a friend who actually knows me, follows my Instagram, and chats with me about my interests.
I asked for simple recommendations to begin.
Without Fabric MCP, I got generic responses.
With Fabric MCP, Claude picked up:
My history and biographies streak, plus the Mongol rabbit hole, so The Silk Roads landed at the top.
Movie picks that match the same theme, including Oppenheimer and Napoleon, with a nod to my international background and theatre habit.
A couple of books I have already read. That is expected because Fabric is not connected to my Amazon account yet.
Several other interests you can see in the video.
Test 2: Finding my trip and preparing an itinerary
I made this one confounding on purpose. I have a few short trips coming up, and I also search for places I am not booking yet, like Mongolia.
With Fabric MCP, Claude correctly identified Capri as the actual upcoming trip and wrote an itinerary I will use since I have not planned it yet. The full itinerary is in the video.
The plan matched my taste in restaurants, leaned into the history I care about, and even recognised one of my main reasons for choosing Capri: shopping!
Test 3: knowledge of specific personal events
I wanted to see if Claude could handle very specific questions that require Fabric MCP’s richer signals, including our asset descriptions for images and videos.
Question: What did I do this morning?
On Sunday morning, I rescued a tiny bird on my way to the office and posted the clips on Instagram.
With Fabric MCP, Claude:
Reconstructed my morning adventure as a clean, ordered timeline.
Correctly identified the bird as a goldcrest.
Even called out the Sunday fire alarm test in the background.
I tried to confuse it by asking about last night without timestamps. I knew I had posted a story after midnight.
Claude still got it. It recognized the wine bar setting, clocked the Barolo on the table, and even called out my outfit and the bag I was carrying.
It tied the late night to the early morning bird rescue and correctly inferred that I did not sleep much.
Test 5: Summarise my interests
As a final pass, I asked for a summary of my current interests.
Using Fabric MCP, Claude:
Captured my career path, skills, and past employers.
Pulled history themes from my books and podcasts.
Remembered my recent theatre visits to Hamilton and Les Misérables.
Read my style signals, named specific brands and items, and even noted my plan to restore a few pieces.
Called my wardrobe color palette: black, white, red.
Listed upcoming trips, my gym (Third Space), supplements I have been researching, and even niche interests like VR.
Why this matters
When AI products start from lived context, they stop guessing. You prompt less, you correct less, and the replies are tailored to you.
You choose what to connect, who can use it, and for what. Nothing happens without your permission.
When your context can travel to the tools you actually use, travel plans, shopping recommendations, and everyday responses feel magical from day one.
Your context can make your AI experiences magical. Fabric MCP is how that context becomes something your models can access and use.