Help & FAQ

Fundaigo is a free AI-powered grant intelligence platform for small nonprofits in education. It runs a 5-layer pipeline, discovery, link check, qualification scoring, bias guard, and tracking, so your team spends 30 minutes a week instead of hours finding scored funding leads. Always confirm a funder's link yourself before applying.

(1) A free Groq API key from console.groq.com, no credit card. (2) A Google account, for NotebookLM and a Shared Drive to hold your team's tracker. (3) Your org's core documents for NotebookLM (mission statement, program descriptions, 501(c)(3) letter, impact stats). Setup takes under 30 minutes.

Go to console.groq.com → sign up (free, no credit card) → click API Keys in the sidebar → click Create API Key → name it Fundaigo → copy the key. It starts with gsk_. Takes about 3 minutes total.

Yes, genuinely free. No credit card required to start. Fundaigo runs on Groq's free API tier, 1,000 requests and 200,000 tokens per day, resetting every 24 hours. One weekly full scan uses under 10% of that daily limit. No trial period, no expiry date.

Fundaigo uses groq/compound, a model with built-in live web search, to scan the open web across 6 funder categories simultaneously. It runs up to 10 internal searches per call to find a match, then returns a single strongest candidate matching your org profile. That one result then passes through the remaining 4 pipeline layers before you see it.

Currently bypassed: automated link checking was dropping valid URLs on the free tier, so Layer 1.5 is temporarily disabled. Every result is marked "needs check" and passes through to scoring as-is, nothing is dropped for a bad link right now. Always visit the link yourself before applying, some may redirect or 404. Filtering resumes once this is re-enabled.

Every result passes through Layer 2.5 which checks four things: language quality, demographic bias, factual accuracy, and nonprofit appropriateness. A ⚠ Guard flagged badge means at least one check raised a concern. The result is shown with a warning rather than hidden, your staff reviews and decides. AI surfaces the flag; humans make the final call. Always.

By design, each run surfaces one candidate opportunity from Layer 1. That single result then passes through a link check (currently bypassed, see the FAQ above), qualification scoring, and the guard check. At the end of the run, it's either dropped, if it scores below the fit threshold, or it reaches you as a scored lead flagged for your review and decision, confirm the link yourself before applying. One result per run keeps each review focused. Run the search again for the next candidate.

Every grant opportunity is scored 1-5 by the Qualification Agent against your org profile. The rubric is explicit, the AI applies the same criteria every run:

5 · Perfect fit: Mission directly matches the funder's stated focus. Location explicitly eligible. Amount within your stated grant range. No special requirements or flags.

4 · Strong fit: Mission closely aligns. Location eligible. Amount within range. Minor manageable flags such as a Letter of Intent required.

3 · Moderate fit: Mission partially aligns or is an adjacent focus area. Location eligible but not the funder's primary geography. Amount slightly outside your range.

2 · Weak fit: Mission alignment is a stretch. Location may need verification. Amount significantly outside range. Multiple flags to investigate.

1 · Poor fit: Mission mismatch, funder focus is unrelated to your work. Location ineligible. Amount far outside range. Major barriers such as invitation-only or school district sponsorship required.

If the result scores below 3, it's dropped rather than shown to you as an active opportunity.

Once a week is the recommended cadence. The memory layer automatically excludes funders already found in previous runs, so each weekly search surfaces genuinely new leads. Set a 30-minute calendar reminder on the same day each week and make it a routine.

Six categories: Community foundations (local and regional, rolling deadlines), Corporate giving (company programs, often unpublicized), Government grants (federal, state, Title I-adjacent), Faith-based funders (churches, dioceses, religious foundations), Family foundations (private, rarely publicized), and Healthcare foundations (health equity, trauma-informed programs). The "Full scan" chip searches all six in one run.

Yes. All tracker data is saved to localStorage, it survives closing the tab, closing the browser, and restarting your computer. It is not the same as browser cache. Clearing your browser cache will not delete it. Only explicitly clearing "Cookies and site data" in your browser's privacy settings would remove it.

Currently the tracker lives in one browser on one device. Use Export CSV to share with your team instantly via Google Sheets. Coming in Phase 2: Google Sheets direct sync so the full tracker is accessible from any device for any team member, without exporting.

When you open the Tracker tab, the system checks every deadline in your pipeline. Any opportunity with a deadline within 14 days triggers a red alert banner at the top of the tracker listing each urgent grant by name, amount, and exact date. Deadline badges on each row also change color: green (>30 days), amber (14-30 days), red (<14 days).

Seven statuses: New (just added), Pursuing (actively working on application), Submitted (application sent), Funded (award received), Declined (rejected this cycle), Not Eligible (confirmed doesn't qualify), Deferred (skipping this cycle, revisit later). Funded and Submitted funders are automatically excluded from future discovery searches.

Google NotebookLM is a free AI tool that answers questions and drafts content based only on documents you upload, it cannot make things up from its training data. Fundaigo uses it as your org's knowledge base. Upload your mission statement, program descriptions, and impact stats once, and every proposal draft draws from your real facts, not invented ones. Free at notebooklm.google.com.

It produces a strong first draft, not a finished, submittable proposal. Fundaigo generates a tailored prompt you paste into NotebookLM alongside the funder's RFP. NotebookLM drafts using your uploaded documents. Staff then fills in [INSERT: ...] placeholders, edits for the org's voice, verifies every statistic, and gives final approval before submission. AI assists. Humans decide and submit. Always.

When NotebookLM needs information that isn't in your uploaded documents, it writes [INSERT: description of what's needed] rather than inventing content. This is a safety feature. Every placeholder must be filled with accurate information before submitting, a funder will reject or disqualify any proposal with unfilled placeholders.

Required: Mission/vision statement, program descriptions, student outcome data and impact statistics, IRS 501(c)(3) determination letter, past funders list.

Recommended: Annual budget overview, student testimonials, geographic service areas (cities and zip codes).

Upload once. The Proposal tab in Fundaigo shows the full checklist with required vs optional labels.

All free under one Groq API key:

Layer 1 + 1.5: groq/compound · live web search; URL link check via visit_website is currently bypassed, see the FAQ above
Layer 2: openai/gpt-oss-120b · qualification scoring 1-5 with reasoning_effort: "low"
Layer 2.5: openai/gpt-oss-safeguard-20b · bias, accuracy, and appropriateness guard
Layer 4: Google NotebookLM · proposal drafting (no API key required)

Fundaigo stores tracker rows, org profile, and search memory locally in your browser via localStorage. There are no Fundaigo servers, no data leaves your device except the search prompts sent to Groq's API. Org documents stay in your Google NotebookLM account. Your Groq API key is stored in sessionStorage only and is never written to disk.

Yes. One line in config.js swaps any model in any pipeline layer. The system is provider-agnostic by design, built on Groq today because it's free with no credit card, but ready for Gemini, Mistral, or any open-source model tomorrow. The pipeline logic is completely separate from the model selection.

Fundaigo shows a clear message: "Rate limit reached. Wait 60 seconds and try again." The per-minute limit resets every 60 seconds. The daily limit resets at midnight UTC. At the recommended weekly usage of one full scan per week, hitting the daily limit is essentially impossible, one scan uses under 10% of the daily allowance.

All five tabs share JavaScript state, tracker rows, pipeline results, and org profile all live in the same memory scope. Splitting into separate HTML files would break cross-tab communication: the Proposal tab needs tracker data, the Discover tab writes to the same memory the Settings tab reads. One shell file, eight separate JS modules for business logic, is the right architecture for vanilla JS with shared state. A React or Vue app would restructure this with a component store.

What it means: The qualification layer (Layer 2) hit Groq's per-minute token limit of 8,000 TPM. This happens when Layer 1 (discovery) and Layer 2 (qualification) both run in the same 60-second window, combining to exceed the limit.

What to do: Wait 65 seconds and click Run search again. The TPM limit resets every 60 seconds automatically. The system will retry and complete successfully.

Why it happens: Groq's free tier shares an 8,000 token-per-minute limit across all models. Layer 1 uses ~3,500 tokens, and Layer 2 immediately requests ~5,000 more, together exceeding the limit.

Long-term fix: The system adds a 65-second wait between layers automatically on retry. If you see this repeatedly, wait 2 minutes between searches.

What it means: The guard agent (Layer 2.5) ran successfully but its response wasn't in the expected format, so the system couldn't automatically read the pass/flag decision. This is a soft warning, not an error. The opportunity still appears in results.

What to do: Review the opportunity yourself before adding it to the tracker. Check that the funder is real, the URL works, and the grant is open. This is exactly what staff should do regardless of what the guard says.

Why it shows: By design, Fundaigo never silently passes something it can't fully assess. The warning is the system being honest, AI flags uncertainty, human decides.

What it means: The combined size of the system prompt + org profile + exclude list exceeded Groq's request size limit.

What to do:

• Shorten your org mission in Settings (keep to 1-2 sentences)
• Shorten your focus areas (3-5 keywords, not full sentences)
• Go to Settings → Reset all memory to clear the exclude list

Why it happens: The exclude list grows with every run. The system caps it at 20 funders automatically, but a very long org profile can push the request over the limit.

What it means: The Groq API key entered in Settings is invalid, expired, or was not saved correctly.

What to do:

• Go to Settings tab → check the key starts with gsk_
• Re-paste the key and click Save key again
• If the key is new, wait 30 seconds after creating it before using
• Get a new key at console.groq.com if needed

Note: Keys are stored in session memory only, they clear when you close the tab. Re-enter your key at the start of each session.

What it means: The pipeline ran but the discovered candidate was dropped before reaching you, it scored below the minimum fit threshold.

What to do:

• Wait 2 minutes and run again, a different search result may come back
• Try a different search category button instead of the default
• Type a more specific custom search in the text box
• Check your org profile in Settings, a very narrow profile can cause all results to fail the fit score
• Go to Settings → Reset all memory if the exclude list has grown large

What it means: The request to Groq's API failed, either a network issue, a Groq service outage, or the request timed out.

What to do:

• Check your internet connection
• Wait 30 seconds and try again
• Check status.groq.com for any outages
• If using a VPN, try disabling it, some VPNs block API requests

Phase 2: Google Sheets direct sync, Google Sign-In via OAuth, automated weekly email digest of new opportunities.

Phase 3: Multi-org SaaS platform on Netlify + Supabase, each nonprofit gets isolated profile, tracker, and memory. Free for small nonprofits under $500K budget.

Phase 4: AI router, config.js intelligently selects the optimal free model per pipeline task based on available keys and org preference (speed vs accuracy).

Fundaigo runs on free tools by design. The underlying AI providers (Groq) currently offer free tiers sufficient for weekly nonprofit use, but LLM pricing is a market decision outside our control. If a provider changes pricing, the one-line model swap in config.js lets any org move to a different free provider immediately. The system is designed to never depend on a single vendor.