Search-grounded funding research

Grant Scout finds the right funding for your mission.

Live search over SerpApi and grants.gov, reasoned by Gemini - every program linked to its real source.

  • Live SerpApi and grants.gov results, not training-data recall
  • Every mentioned program links to its real search result
  • No incorporation required to start looking

Try it

Ask Grant Scout a funding question.

SerpApi + grants.gov → Gemini · cited sources

Answers come from a live search run when you press send, with sources listed below the answer. Search history is off by default and this demo doesn't save or publish your question.

How it works

Three steps, and none of them involve guessing.

The agent is deliberately narrow: retrieve real results, reason over them, show its work. If a program can't be traced to a live page, it doesn't make the list.

  1. 01

    Ask in plain language

    Describe the money you need and the shape you're in - solo, unincorporated, pre-revenue, side project. No forms, no eligibility quiz, no jargon required.

  2. 02

    A live search runs, not a memory lookup

    For US federal questions, Grant Scout calls grants.gov's own Search API for real opportunity data; otherwise it calls Google Search through SerpApi. Either way, the call happens at request time, so closed rounds and stale deadlines don't get repeated back as fact.

  3. 03

    Gemini flags potential eligibility

    The reasoning agent reviews the returned titles and snippets, flags programs that may require an incorporated entity, and links each mentioned program to its search result for your verification.

Verified authority

Two live sources. No footnotes pretending to be facts.

A cited source is only worth trusting if it was fetched when you asked, not recalled from training data. Every answer traces back to one of these two calls.

SOURCE: GRANTS.GOV

Federal opportunity search API

Every US federal question hits grants.gov's own Search API directly - the same opportunity data agencies publish, not a cached summary of it.

queried live, at request time - not cached

SOURCE: LIVE WEB SEARCH

SerpApi-backed Google Search

Non-federal and program-specific questions run a live Google Search through SerpApi. Tracking-redirect links are filtered out - every citation is the real destination URL.

queried live, at request time - not cached

Go beyond the basics

You're not just asking a question. You're commanding a research analyst.

Typing a plain search is the entry point, not the ceiling. Here's how people who actually push this thing get more out of every question.

Stack two programs in one ask

Don't run two searches. Name both programs in a single question and Scout tells you where they overlap - and where they conflict.

Compare NSF SBIR Phase I eligibility to California's state matching-grant programs for the same project.

Audit a lead you found elsewhere

Saw a program mentioned on a forum, in a newsletter, by a friend? Paste the name back to Scout and get it verified against a live source, not vibes.

Is the program someone called the "Rural Innovation Fund" still accepting applications right now?

Hunt by deadline, not by name

You don't need to know what a program is called to find it. Ask what's closing soon and let the live search surface it.

What federal grants for a solo AI founder close in the next 30 days?

Describe the problem, not the program

Skip the grant-speak entirely. Describe what you actually built and let Scout map it to categories you didn't know you qualified for.

I built a free curriculum app for rural schools with zero budget - what funding categories should I even be looking at?

Try one of these on the agent up top

Beyond the demo

The agent pattern

Grant Scout is the narrowest possible slice of a general pattern: retrieve real results at request time, reason over only what came back, cite the exact source for every claim, and refuse to answer when nothing verifiable turned up. Swap the domain and the source APIs, and the pipeline in _lib.py doesn't change shape.

Public benefits eligibility

Same sort: retrieve a live benefits-program page for SNAP, housing, or utility assistance, then reason over the actual eligibility text instead of the stale advice on a forum thread.

Scholarships for the institution-less

Grant Scout sorts "needs an LLC" vs "doesn't." The identical filter works for a student with no university affiliation searching fellowships that assume one.

Business license & permit lookups

Swap grants.gov's search2 endpoint for a city or state permitting API. Same shape: live query, structured hits, a link to the actual ordinance page - not a remembered rule that changed last year.

Vendor & certification checks

Is this supplier still on the exclusion list? Still licensed? A retrieve-then-cite agent answers from the live registry at request time, where a cached answer is already wrong the day it's stale.

Public procurement discovery

The same routing logic that sends a federal question to grants.gov instead of SerpApi extends cleanly to RFPs and government contract postings - a different live source, the identical pipeline.

Now / next - what's honestly not built yet

TodayNext
One question, one answer - no memory of what you asked beforeMulti-turn refinement that keeps prior sources in context without re-searching
The tool fires exactly once per request, by hard rule, to protect latencyThe agent decides when a second, narrower search is actually worth the wait
You have to come back and ask again to see what's newA saved query re-runs on a schedule and flags newly posted matches
grants.gov vs. SerpApi is a fixed keyword list, not a model decisionThe model routes itself, once its two-tool judgment holds up under real testing
Eligibility is reasoned from the snippet alone - the linked page is never openedA verification pass fetches the real page before the final answer ships

This is just the beginning

See the code. Fork it. Make it yours.

Two tools, one agent, a hard rule against answering without a live source. Everything above traces to real code, not a pitch deck - go read it.