TikTok videos about insulated tumblers with over 100k views, most viewed first
- “Ice still there after 24 h”
CROSS-CHANNEL · DATA SEARCH
Research across channels, without the tab shuffle
Start with a plain-language question about products, videos, creators or posts, then follow supported links and check each row's evidence.
- Keble accounts are invite-only for now
Their creators · served link
- @sample.sips
Creators like @sample.sipsSimilar
Several tools, then another spreadsheet
One product idea, researched by hand:
- Products in a marketplace tool
Prices and reviews in one tab.
- Videos in another app
Scroll to see what gets views.
- Creators copied by hand
Handles pasted into a sheet.
- Forum threads, one by one
Complaints read post by post.
Start with one question, keep investigating
Each question runs on its own. An arrow means Keble serves that link.
Sample data- Recorded, mentioned or similar
Every row says how Keble knows it.
RecordedMentionedSimilar - Unapplied conditions are listed
Words the search could not use are shown, not dropped.
Not applied: “dishwasher safe” - An observed time on every row
You see how old each row is.
Observed 2 h ago
Hand the same search to your AI Agent
Connect Keble over MCP and paste the prompt. 1 request per search, in Keble or from your AI Agent.
US kitchen gadgets under $30 with the most reviews
Keble · data search✓Found US kitchen gadgets under $30, most reviews first. Each row says recorded or similar, with its observed time.
View prompt
Use the Keble MCP tools. First call dp_catalog to see which kinds and links are served. Then call dp_search with the question "US kitchen gadgets under $30 with the most reviews" and set hints.market to US. For each result, say how Keble knows it (recorded, mentioned or similar) and when it was observed, and list any words the search could not apply. Follow the cursor with dp_continue until you have 20 rows, open the top 3 with dp_record, and give me a table with name, price, review count, observed time and evidence.
Questions
What does a search cost?
Each search counts as 1 request, and you see that before it runs.
Where does the data come from?
Public marketplace, short-video and forum data that Keble collects and analyses.
How fresh is it?
Each row shows when it was observed, and the coverage table dates each source.
Do examples run a search?
No. It only fills the box; nothing runs until you press search.
Start your research with one question
- Keble accounts are invite-only for now