Grant Scout

What it does

Grant Scout helps solo, non-incorporated founders find funding programs they can actually apply for. Ask a plain-language question - "grants for a non-incorporated solo founder building an AI agent product in the US" - and Grant Scout runs a live Google Search through SerpApi, or - for clearly US federal questions - a live search against grants.gov's own public Search API for real opportunity data (real opportunity number, agency, close date, and a real grants.gov detail-page link). Gemini reviews only the real returned results and explains which programs plausibly don't require a legal entity, citing the real source URL for every claim. Nothing is fabricated: if a program doesn't appear in a real search result, it never makes the list.

We built this because it's a real problem for our own team. ArkNet Digital is a real, currently non-incorporated solo-founder operation, and finding non-dilutive funding that doesn't assume you already have an LLC or C-corp is a genuine, unsolved friction point. Grant Scout is the tool we wished existed.

Why this could be a real business, not just a demo

The addressable user isn't hypothetical - solo/indie AI builders are one of the fastest-growing segments on platforms like Devpost, Product Hunt, and Indie Hackers, and nearly all of them face the same non-incorporated-founder funding gap this tool was built to close. The workaround today is manually reading eligibility fine print across dozens of grant pages, or giving up and assuming "grants are for companies" - a real, specific, and currently unsolved piece of friction, not a nice-to-have.

Path to a real product, honestly scoped for where this is today:

• This hackathon build is a working prototype, not a monetized product yet - that's the honest state.
• The natural monetization shape is subscription (continuous monitoring + alerts when a founder's specific eligibility profile matches a newly posted program) rather than pay-per-search, since the recurring value is in not having to re-search every few weeks.
• Distribution is realistic without paid acquisition: indie-builder communities (Indie Hackers, Product Hunt launches, X/Twitter's build-in-public crowd, Devpost itself) are exactly where non-incorporated solo founders already look for this kind of tool, and word-of-mouth in that community is dense.
• Defensibility isn't the search itself - it's the eligibility-reasoning layer (grants.gov + SerpApi results filtered specifically for "no entity required") that a generic search or ChatGPT session won't give a founder without real digging.
• Kill criterion we'd actually apply: if a small cohort of real non-incorporated founders don't return to re-run a search within a few weeks of first use, the recurring-value thesis is wrong and this stays a one-off tool, not a company.

How it works

Agent: Built on Google's Agent Development Kit (google-adk), running a single Agent with two live-grounding tools - tool_search_funding_programs and tool_search_grants_gov. Live grounding: tool_search_funding_programs calls SerpApi's Google Search API (https://serpapi.com/search) at request time; tool_search_grants_gov calls grants.gov's own public Search API (https://api.grants.gov/v1/api/search2, no API key required) for real US federal opportunities. Which tool runs is decided deterministically before the agent is even built, based on whether the question signals US federal funding (mentions "federal", "grants.gov", SBIR/STTR, a specific federal agency) - never left to model judgment, since testing found the model didn't reliably choose the more precise source on its own. Reasoning: the agent (gemini-flash-lite-latest via the free Gemini Developer API) reviews the real results returned by whichever tool ran, flagging which ones plausibly fit a non-incorporated applicant and which likely require a registered entity - always citing its source. Backend: the same agent code runs inside a Vercel Python serverless function (frontend/api/chat.py), avoiding a Python framework preset so it coexists with the Next.js frontend in one deployment. Frontend: Next.js 16 (App Router), React 19, Tailwind CSS v4 - a trustworthy, financial-adjacent design (navy-to-emerald gradient, Inter type).

SerpApi + grants.gov integration

Neither is a decorative add-on - together they're the entire grounding layer. Every single answer the agent gives is the direct result of a live request made in that request, to whichever source is the more authoritative fit for the question. There is no fallback to the model's own training data; if the chosen tool returns no usable results, the other tool is tried once as a fallback, and if both come up empty the agent says so rather than guessing.

Try it live

https://grant-scout-jayblast.vercel.app - ask it a real funding question and watch it run a live search in front of you, with cited sources underneath the answer.

Built With

• gemini
• google-adk
• nextjs
• python
• react
• serpapi
• tailwindcss
• typescript
• vercel