✦ Case study

Every funding call worth your time — found and scored for you.

Their fundraiser hunted dozens of platforms and Telegram channels by hand. Now an agent reads every source on a schedule, scores each grant against their own criteria, and surfaces only the ones worth reading — with the reasoning attached.

Client: GGTC Kyiv
Industry: Grant fundraising · GovTech
Engagement: Custom grant-discovery agent on our infrastructure
Timeline: 5 days, kick-off to live

✦ Challenge

The problem we were brought in to solve.

GGTC Kyiv raises money through grants — and finding the right calls meant one fundraising expert working through dozens of platforms, embassy pages and Telegram channels by hand, week after week. The sources that matter most to them — Ukrainian and EU funding portals, ministry PDFs, informal Telegram feeds — are exactly the ones no off-the-shelf grant database indexes.

The work was slow, easy to fall behind on, and impossible to hand off: what makes a grant worth chasing — the right donor, a budget above $50,000, an eligible geography, a realistic deadline — lived in her head, not on paper. Miss a call and a six-figure opportunity is gone with its deadline.

BeforeWith Granty

One expert working through dozens of platforms and Telegram channels by hand, over and over

An agent reads every source on a schedule — nights and weekends included

“Is this grant worth it?” judged from memory, one at a time

Every call scored 0–10 against written criteria, with the reasoning shown

What counts as a good grant lived in one person's head

The criteria live in GGTC's own Google Doc — edited anytime, no developer

Skim a long list and hope nothing important slipped through

420 opportunities filtered down to the 28 worth a human's time

✦ What we built

What we built and how it runs.

We built Granty — a grant-discovery agent that runs on our infrastructure and reads the sources GGTC already trusts, then scores what it finds against the client's own written criteria.

It runs on a schedule or on demand, hands the strong matches to a dashboard the team logs into, and mirrors them to a shared Google Sheet. Every recommendation is a first pass, never the final call — a human decides.

GGTC's fundraiser

Keeps the source list and the criteria in their own Google files — no tickets, no developer.

Every source GGTC trusts

Grantyreads each sourcescores 0–10recommends ≥7✓ shortlist

Grant platformsTelegram channelsEU Funding & Tenders APIEmbassy & ministry pagesTheir own criteria Doc

✦Granty reads the informal sources structured grant databases skip — Telegram, ministry PDFs, EU calls — and scores them against GGTC's own written criteria.

Scores against their criteria

The rubric lives in GGTC's own Google Doc. Edit it, hit “Rescan”, and the agent shows a diff — green added, red removed — before anything changes.

criteria: a client-editable Doc

Never a black box

Every score opens to a breakdown across seven criteria — mission, budget, eligibility, geography, donor, deadline, consortium — each with a written reason it fits.

7 criteria, shown per grant

Reads what databases skip

Telegram channels, EU calls via the official API, JS-heavy portals via a headless browser — the informal UA/EU sources no grant database indexes.

Telegram · EU API · headless

What Granty takes off their plate

→Scan every platform and Telegram channel on a schedule
→Score each opportunity against your own written criteria
→See exactly why a grant scored what it did
→Edit the criteria yourself — anytime
→Push the strong matches to a shared Google Sheet

StackOpenRouter · GeminiGoogle SheetsGoogle DocsTelegramNext.js

✦ Under the hood

What happens in between.

1
Read
On a schedule or one click, the agent pulls every source from GGTC's Google Sheet and fetches each one.
~35 sources · every 5 days
2
Extract
It turns each page — HTML, Telegram, EU portal, even JS-only sites — into a clean list of funding opportunities.
listings → opportunities
3
Score
Each opportunity is scored 0–10 against the client's written criteria, with a reason for every criterion.
0–10 · per-criterion
4
Shortlist
Only scores ≥7 are recommended. The rest stay searchable but out of the way.
≥7 recommended
5
Review
The fundraiser triages the shortlist — reviewed or rejected — and those notes calibrate future scoring.
human decides

✦ The result

The service GGTC logs into.

The GGTC Grant Agent dashboard — candidate count and controls — The home screen: how many opportunities the agent scanned, how many it recommends, and one button to run it now or let it run on a schedule.

The shortlist of scored grant opportunities — The shortlist. Every row is a funding call with a 0–10 score, budget, deadline and donor — sorted so the strongest sit on top.

A single grant's score broken down by criterion — Open any row and the score breaks down across all seven criteria, each with a written reason. Never a black box.

✦ Results

What it changed.

28 / 420

Worth reading, of everything scanned

The agent scores every opportunity it finds and recommends only those ≥7. The team reviews a shortlist, not a firehose.

≈ $2/mo

To keep it running

One full scan of every source costs about $0.20 in API. Running it every few days is a couple of dollars a month.

~35

Sources, every run

Platforms, embassy pages, EU portals and Telegram channels — every source on GGTC's list.

✦For context: structured grant databases like Instrumentl or Candid don't index Ukrainian portals, ministry PDFs or Telegram feeds — the exact sources GGTC relies on. Granty was built around them.

Granty

Grant Scout

Employee file