10 TinySet Use Cases: Turn Web Data into a Spreadsheet from One Sentence

Have a table you keep building by hand? Try TinySet free. Describe the rows you want in one sentence and get a sourced table back.

TL;DR

Getting web data into a spreadsheet usually means one of two things: copy and paste from a dozen tabs, or build a scraper for each site. TinySet is a third option. You describe the table in one sentence, check the columns it proposes, and it searches the public web for matching companies, products, places or programs and fills one row per entity, with values taken from real pages.

This post covers 10 practical TinySet use cases, each with who it's for, a prompt you can copy, the columns to ask for and a tip. Three of them I ran for real while writing this, and I show the actual rows, gaps included. It also covers how to run TinySet from the playground, ChatGPT and Claude, and the API, MCP or CLI.

A quick disclosure: I work at TinyFish, which makes TinySet. Where something didn't fill or looked off in my runs, I say so.

What TinySet is, in one paragraph

TinySet builds a structured dataset from a plain-English description. You say what each row is, the scope, and the fields you need. It plans typed columns, discovers matching entities across the public web, and collects each field from pages it actually read. You can ask for up to 250 rows per build and up to 64 columns. A build usually takes 5 to 20 minutes, and each complete row costs $0.10 from your TinyFish Wallet. When it's done, you can export the table, share a read-only link, or read the rows from code.

It's one part of TinyFish, which is one API with one key. Use TinySet when you know the fields and need the same facts across many entities. Use TinyResearch when you want a written, cited report instead of a table, and TinyMonitor when you want to watch a page or topic for changes.

What people use TinySet for today

Since TinySet launched on November 5, 2026, most tables people build fall into a handful of jobs:

• Market maps: every company in a category, with what it does, pricing and website.
• Prospect and account lists: companies that fit a profile, ready to drop into a CRM.
• Events and opportunities: conferences, hackathons, festivals and calls for applications, with dates and deadlines.
• Grants, programs and funding: fellowships, accelerators and credit programs, with eligibility and how to apply.
• Supplier and product comparisons: vendors or tools side by side, with prices, features and limits.
• Local directories: businesses or venues in one city, with addresses, hours and contact pages.

The 10 use cases below follow those jobs, each with a prompt you can copy.

How to use TinySet

You can run TinySet from wherever you already work. All of these use the same TinyFish account.

In the playground. Open the Datasets playground, type one sentence and click Build. TinySet plans the columns, which can take up to a minute. Review them: rename, change types, pick the primary key that decides when two rows are the same entity, add a filter on a yes/no column, and set the row count from 1 to 250. Then click Build dataset and watch rows appear. Our step-by-step guide, Build a Dataset with AI Using TinySet, walks through every screen.

In ChatGPT or Claude. Install the TinyFish plugin and ask for a table in plain language, for example "build me a dataset of coworking spaces in Lisbon with a day pass, price and website". The assistant calls the dataset tools for you and shows the rows when they're ready.

From code, MCP or the CLI. The flow is the same everywhere:

1. plan_dataset turns a prompt into a spec you can edit.
2. create_dataset starts the build from a prompt or a spec, with a target_rows value.
3. get_dataset checks status and returns the first rows; get_dataset_rows pages through the rest.
4. populate_dataset fills an existing dataset again, to add rows or rebuild it.

In Python, a full build is a few lines:

from tinyfish import TinyFish

client = TinyFish()  # reads TINYFISH_API_KEY
dataset = client.datasets.create(
    prompt="Specialty coffee shops in Oakland with their website and rating",
    target_rows=25,
)
page = client.datasets.rows(dataset.id)  # once the run is ready
for row in page.rows:
    print(row.data)

Or from the terminal: tinyfish dataset create "Specialty coffee shops in Oakland with their website and rating" --rows 25 --pretty. The Datasets docs have the full reference, including the run event stream.

Three real builds

I ran three of the use cases below through TinyFish on November 9, 2026, starting at about 10:59 PM PT, at 10 rows each. Here's what came back, unedited.

Build 1: AI coding assistants compared

Prompt: "AI coding assistants, with the company that makes it, the cheapest paid plan price per month in USD, whether there is a free tier, the IDEs or editors it supports, and the pricing page URL"

TinySet planned six columns: product (the key), company, cheapest paid plan, free tier, supported IDEs and pricing page URL. All 10 rows came back in about a minute and a half. A few of them:

Two honest notes. Where TinySet couldn't confirm a value, it left the cell blank instead of guessing, which is what you want in a comparison. And pricing pages change often, so treat the prices as a snapshot from the night I ran it, and click through to the pricing page before you quote one.

Build 2: Festivals in Tasmania that take performers or stallholders

Prompt: "Festivals and community events in Tasmania, Australia, taking place in 2027, with event dates, town, organiser, official website, and a link to the page where performers, entertainers or stallholders can apply"

This one is inspired by a real TinyFish user. Shields Events, a magic and entertainment business in regional Tasmania, uses TinyFish to find festivals and council events to apply to, instead of trawling event pages one by one. TinySet planned seven columns and found 10 events, including:

Four of the 10 events had an application page TinySet could find, and the organiser column had gaps too. That's the useful part: the blanks tell you exactly which festivals need an email instead of a form.

Build 3: Grants and fellowships for AI builders

Prompt: "Grants and fellowships for AI builders and researchers that are open to applicants in 2027, with the funding organisation, award amount, who is eligible, the application deadline, and the official application page URL"

All 10 rows came back, each with an application link. A few:

This build also shows why you review the result. One row listed an award of "$550 minimum", which is almost certainly a misread figure, and one row was a general NIH grant type rather than a specific AI program. A good prompt gets you a strong first draft; a quick review turns it into something you'd send to someone.

Want to try one of these yourself? Build your first dataset free. New TinyFish accounts start with $8 in Wallet funds, enough for a first build.

10 TinySet use cases

1. Map a market or category

Who it's for: founders, product marketers, investors, strategy teams.

Prompt: "AI inference providers with company, headquarters, supported open-weight models, cheapest price per million tokens and pricing page URL"

Columns to ask for: product, company, positioning line, key feature, lowest public price, pricing page, last funding round.

Tip: ask for a source URL column for anything you'll quote. Market maps get shared, and a link per row saves you the "where did this number come from?" thread later.

2. Build a pricing battlecard

Who it's for: product marketing, sales enablement.

Prompt: "CRM tools for small sales teams with vendor, cheapest paid plan per user per month, free plan yes or no, seat minimum and pricing page URL"

Columns to ask for: product, plan name, price, billing period, free plan, limits, pricing page.

Tip: once the table is right, set a TinyMonitor on each pricing page so you hear when a competitor changes a price, instead of rebuilding the table. We cover the monitoring side in competitor price tracking.

3. Find events and opportunities to apply to

Who it's for: small businesses, performers, makers, community organisers.

Prompt: see Build 2 above. Swap the region and the kind of event: "craft and makers markets in Victoria in 2027 with dates, organiser and stallholder application link".

Columns to ask for: event, dates, town, organiser, website, application link, application deadline.

Tip: add an application-deadline column and sort by it. The blanks in the apply-link column become your email list for the week.

4. Track grants, fellowships and non-dilutive funding

Who it's for: researchers, students, founders.

Prompt: see Build 3 above, or narrow it: "non-dilutive grants for climate tech startups in the UK open in 2027 with funder, amount, eligibility, deadline and application page".

Columns to ask for: program, funder, amount, eligibility, deadline, application page, region.

Tip: deadlines move. Re-run the build each quarter with Add rows or a rebuild, and check each deadline on the official page before you plan around it.

5. Build a target account dataset

Who it's for: founders doing early sales, B2B sales and marketing teams.

Prompt: "B2B SaaS companies in Europe that sell to logistics firms, with website, headquarters city, product category, public pricing yes or no and careers page URL"

Columns to ask for: company, website, country, what they sell, company size signal, public pricing, careers page.

Tip: keep this at the company level and use public business pages. Then use TinyResearch on your top 10 rows for a cited brief on each account before you reach out.

6. Source suppliers and vendors

Who it's for: e-commerce founders, procurement, operations.

Prompt: "Manufacturers of stainless steel water bottles that export to Europe, with company, country, certifications, minimum order quantity, website and public sales contact page"

Columns to ask for: supplier, country, product range, certifications, MOQ, lead time, website.

Tip: expect fields like MOQ and lead time to fill less often, because many suppliers only share them on request. Ask for them anyway; the gaps tell you who to contact first.

7. Build a reference dataset of projects or facilities

Who it's for: analysts, researchers, operations and infrastructure teams.

Prompt: "US data center projects under construction with operator, city, state, planned capacity in MW, project status and source URL"

Columns to ask for: project, operator, location, capacity, status, announcement date, source URL.

Tip: this is one of our own launch examples. A 100-row build returned 99 rows with a source per project. Pick a stable primary key, such as project name plus city, so two phases of the same campus don't collapse into one row.

8. Make a local business directory

Who it's for: agencies, freelancers, local service businesses, city guides.

Prompt: "Independent dental practices in Austin with website, neighborhood, phone and the public contact email published on their site"

Columns to ask for: business, website, neighborhood, phone, public contact email, opening hours, rating.

Tip: use the website as the primary key, not the name. Two locations of the same practice share a name but not a page. And only ask for contact details a business publishes for that purpose.

9. Track hiring and programs

Who it's for: students, job seekers, recruiters, career services.

Prompt: "AI companies in New York with an open software engineering internship for summer 2027, with company, role title, application deadline and careers page URL"

Columns to ask for: company, role, location, deadline, careers page, program type.

Tip: internship pages open and close fast. Pair the table with a TinyMonitor on the careers pages you care most about. A student built a version of this idea as an app; read how they built an AI internship finder that checks careers pages.

10. Feed typed rows into an agent or workflow

Who it's for: developers, data teams, AI builders.

Prompt: "Large language model APIs with provider, model name, context window in tokens, input price per million tokens, output price per million tokens and docs URL"

Columns to ask for: typed fields your code needs: numbers as number, dates as date, URLs as url.

Tip: call plan_dataset first and commit the spec to your repo, then create from the spec. That keeps column names stable between runs, so the code that reads the rows doesn't break. If you work in ChatGPT, you can wrap the whole flow in a skill: see ChatGPT skills.

How to write a TinySet prompt that fills well

From the docs and from what I saw in these builds:

• Name the entity. "Coffee shops" works better than "coffee in Oakland". Each row should be one thing.
• Scope it. A place, a date range or a category: "in Tasmania", "open in 2027", "Series A only".
• List the fields. Every field you name becomes a column. Fields you leave out are left out.
• Prefer facts on public pages. Websites, addresses, prices and dates fill well. Private, login-only or paywalled data doesn't.
• Don't ask for a ranking. "Top 10" returns 10 matching rows, not the 10 best. Ask for a column such as a rating or a price and sort it yourself.
• Start small. 10 to 25 rows shows you whether the columns are right before you pay for 250.

What TinySet is not for

• It's a snapshot, not a feed. A dataset reflects the pages it read during the build. You can add rows or rebuild, but it doesn't refresh on its own. Pair it with TinyMonitor for anything that changes often.
• It's not a census. A row target is a ceiling, not a promise of complete coverage. A finished build can return fewer rows than you asked for.
• It's not for personal data or closed platforms. Use it for public business, product, place and program information. Don't use it to collect personal details about private individuals, and respect each site's terms.
• It's not a written analysis. If you want an argued answer with citations, use TinyResearch. If you need to click through a site or fill a form, use TinyAgent.

Keep reading

• How TinySet works: building a dataset from a simple prompt
• Introducing TinyResearch and TinySet
• Buy vs build web datasets
• Best structured data extraction tools
• AI tools for startups

Turn your next "can someone make a spreadsheet of..." into one sentence. Start free with TinyFish: new accounts get $8 in Wallet funds, and TinySearch and TinyFetch are free within their limits.