Connect your saves to Claude and ChatGPT
Saving is the easy part. You've got hundreds of posts filed away — restaurants, recipes, workouts, places you meant to go. The hard part is that revisiting them requires you to already know what you're looking for.
Search solves that only when you remember the words. It doesn't help when the question is bigger than a keyword — what should I actually do with all this?
What changes when an assistant can read your saves
Willa connects to Claude and ChatGPT over MCP, which means your assistant can query your own library directly. Not the internet at large. The specific posts you bookmarked, with your notes and boards intact.
The difference is that you stop retrieving individual saves and start asking questions of the whole pile.
Prompts that actually work
"Plan three days in Lisbon from what I saved"
Pulls every Lisbon-related save — the reels, the restaurant posts, the neighbourhood guide someone sent you — and builds an itinerary that groups them by area so you're not crossing the city four times.
"What restaurants did I save in Tokyo?"
The question you'd otherwise answer by scrolling. It returns the list with whatever context came along — the dish someone recommended, the note you left about booking ahead.
"Make a shopping list from the recipes I saved this month"
Reads across several saved recipes at once and consolidates the ingredients, rather than making you open each one and write things down twice.
"Which of my saved workouts don't need equipment?"
A filter you never created, applied to saves you never tagged. This is the kind of question that's impossible with folders and trivial with an assistant.
"I saved something about a coffee place in Ubud — what was it?"
Half-remembered retrieval. You don't know the name, which is exactly why keyword search fails and a conversation doesn't.
"Summarise everything I've saved about kitchen renovations"
Useful when a project has been accumulating quietly for months and you've lost track of what you already decided.
What MCP actually is
The Model Context Protocol is an open standard for letting an AI assistant talk to an outside service. Willa runs an MCP server; Claude and ChatGPT support connecting to one.
Two things worth being clear about. Nothing is uploaded or copied into the assistant — it queries your library live, when you ask, and only what your question needs. And it's read access: the assistant looks at your saves, it doesn't reorganise them behind your back.
Setting it up
It's a few clicks. In Claude, open Connectors, add a custom connector, and point it at Willa's MCP URL. ChatGPT follows broadly the same path. You authenticate once with your Willa account and that's it.
The full walkthrough with screenshots for both is on our setup page — it takes about two minutes.
The honest limits
It's only as good as what you've saved. If you bookmarked fifteen reels and never added a note, the assistant has fifteen reels and no notes to work with. The habit still matters.
Willa's AI features are tiered — there's free access with monthly limits, and Premium at $2.99/month removes them. See plans for what's in each.
And we're not the only option. Siloe runs a remote MCP server too, though it sits on their $6.99/month Pro tier and the app is built around bookmarks from text platforms — X, Reddit, LinkedIn — rather than the visual social saves most people accumulate. If your library is mostly threads and articles, look at them seriously.
What's genuinely rare is the combination: saves from anything with a URL, organised into boards, and queryable from the assistant you already have open.