Original Idea
Small Biz Grant Finder A web app that matches grants and assembles a reusable application pack.
Product Requirements Document (PRD): GrantStream
Version: 1.0 (January 2026)
Status: Final / Ready for Sprint Planning
Lead Product Manager: Senior PM
1. Executive Summary
GrantStream is a comprehensive web platform and browser extension designed to eliminate the friction of the small business grant lifecycle. By combining an AI-powered semantic matching engine with a secure, reusable document vault (Reusable Application Pack - RAP), GrantStream allows business owners to discover eligible funding in seconds, manage documentation securely, and generate narrative drafts using multi-agent AI orchestration.
2. Problem Statement
Small business owners and non-profit founders face a fragmented grant landscape. They lose hundreds of hours manually searching for opportunities, repeatedly compiling the same financial documents, and struggling to tailor narratives to specific funder requirements. Current solutions are either static databases or manual consultancy, leaving a gap for an automated, end-to-end "Search-to-Submission" workflow.
3. Goals & Success Metrics
- Discovery Efficiency: Reduce time spent searching for eligible grants by 80%.
- Application Speed: Decrease time-to-submission for repeat applications by 60% through RAP document reuse.
- Success Rate: Increase the grant award rate for users by providing high-quality, AI-critiqued narrative drafts.
- Metrics:
- Monthly Active Users (MAU) exploring >5 grants/month.
- Document Reuse Ratio (number of times a vault document is attached to an application).
- "Match Accuracy" score based on user feedback on recommended grants.
4. User Personas
- Sarah (The Struggling Founder): Owns a boutique tech agency; lacks the time to search for grants and needs a central place for tax/ID docs.
- Marcus (The Non-Profit Director): Needs to track high-volume, low-amount community grants with strict, recurring deadlines.
- David (The Grant Consultant): Manages multiple client profiles and needs a Kanban dashboard to track application progress across various stages.
5. User Stories
- As a business owner, I want to link my bank account via Plaid so that my revenue and employee count are automatically verified for eligibility.
- As an applicant, I want a secure document vault with AES-256 encryption so that I can store sensitive tax returns without fear of data breaches.
- As a power user, I want a browser extension that recognizes when I'm on a grant portal and offers to save that grant to my dashboard.
- As a time-pressed founder, I want an AI agent to draft my "Project Impact" statement based on my business profile and the specific grant criteria.
6. Functional Requirements
6.1. Smart Eligibility Engine
- Hybrid Search: Combine boolean filters (NAICS codes, revenue, location) with semantic vector search (grant objectives).
- Plaid Integration: Automated financial verification (Revenue, Cash Flow) using the "Transactions for Business" API.
6.2. Reusable Application Pack (RAP) Vault
- Document Management: Upload, versioning, and categorization (Tax, Legal, Identity).
- Envelope Encryption: Each document is encrypted with a unique Data Encryption Key (DEK).
- PDF Assembly: Merge multiple documents into a single package with global pagination and a Table of Contents.
6.3. AI Narrative Orchestration
- Multi-Agent Workflow: A "Researcher" agent extracts funder priorities; a "Writer" agent drafts the narrative; an "Auditor" agent checks for compliance.
- Human-in-the-Loop: Users must "nudge" or approve the AI-generated plan before the final draft is produced.
6.4. Application Dashboard
- Kanban View: Visual stages: Draft → Financial Review → Compliance Check → Submitted.
- Deadline Tracking: Automated SMS/Email reminders via Twilio Message Scheduling.
7. Technical Requirements
7.1. Tech Stack (2026 Standards)
- Frontend: Next.js v16.1.3 (Node.js 24 LTS) with Turbopack.
- Styling: Tailwind CSS v4.1.18 (Zero-Config / CSS-First).
- Backend: FastAPI v0.128.0 (Python 3.12+) for AI and PDF services.
- Database: PostgreSQL with pgvectorscale for high-speed semantic matching.
- ORM: Prisma v7.2.0 (TypeScript Engine).
- Auth: Clerk (Satellite Domain configuration for cross-service seamlessness).
7.2. Core Integrations
- Financials: Plaid API (Business Account Indicator).
- Messaging: Twilio SMS & SendGrid Email.
- PDF:
@react-pdf/rendererfor templates andpdf-libfor assembly/merging. - AI: LangGraph for cyclic orchestration of Claude 4 / GPT-5 models.
8. Data Model
8.1. UserProfile
userId(UUID),clerkId(String),email(String).metadata: Industry, Geo, EmployeeCount, Revenue_Verified (Boolean).
8.2. GrantListing (Vectorized)
grantId(UUID),title(String),provider(String).embedding(Vector 1536),deadline(DateTime),criteria(JSON).
8.3. VaultDocument
docId(UUID),userId(FK),docType(Enum: TAX_RETURN, INCORP_DOC).encryptedPath(S3 URL),dekId(KMS Reference),version(Int).
8.4. GrantApplication
appId(UUID),userId(FK),grantId(FK),status(Enum).scheduledReminderId(Twilio SID).
9. API Specification
POST /api/v1/grants/match
- Description: Returns semantically similar grants filtered by hard eligibility.
- Request:
userProfileId,limit. - Response:
List<GrantListing>withmatch_score.
POST /api/v1/vault/assemble
- Description: Merges selected documents into a single PDF.
- Request:
documentIds[],includeTOC(Bool). - Response:
signedUrl(S3) for the generated PDF.
POST /api/v1/ai/narrative
- Description: Triggers multi-agent drafting process.
- Request:
grantId,businessContext,tone. - Response:
narrativeDraft,agentCritique.
10. UI/UX Requirements
- Dashboard: Kanban board using
dnd-kitwith restricted movement (cannot skip "Compliance" stage). - Extension: Chrome Side Panel (Manifest V3) for persistent research context across grant portals.
- Accessibility: WCAG 2.2 compliant "Move To" menus for all Kanban actions (keyboard-first).
- Visuals: Glassmorphism UI using Tailwind 4
backdrop-blur.
11. Non-Functional Requirements
- Security: AES-256-GCM encryption with NIST ML-KEM post-quantum hybrid mode.
- Performance: < 1.5s for hybrid vector search queries.
- Reliability: 99.9% uptime for the Document Vault; 100% notification delivery for T-24hr deadlines.
12. Out of Scope
- Legal advice or direct representation in grant disputes.
- Equity-based funding matching (VC/Angel).
- Direct API submission to government portals (Phase 2).
13. Risks & Mitigations
- AI Hallucination: Mitigated by Agentic RAG (Retrieval-Augmented Generation) and mandatory human "Audit" steps.
- Data Breach: Mitigated by Envelope Encryption and Node.js Permission Model restricting filesystem access.
- Inaccurate Matching: Mitigated by
pgvectorscalepre-filtering on hard metadata before semantic search.
14. Implementation Tasks
Phase 1: Infrastructure & Authentication
- [ ] Initialize Next.js 16.1.3 project with Node.js 24 LTS.
- [ ] Setup Tailwind CSS 4.1.18 using CSS-first configuration.
- [ ] Configure Clerk Auth with Satellite Domain for FastAPI AI backend.
- [ ] Initialize PostgreSQL with
pgvectorandpgvectorscale.
Phase 2: Core Data & Vault
- [ ] Define Prisma 7.2.0 schemas for Users, Grants, and Applications.
- [ ] Implement Document Vault with
AES-256-GCMusing Web Crypto API. - [ ] Build Envelope Encryption module (unique DEK per document).
- [ ] Integrate Plaid "Transactions for Business" for revenue verification.
Phase 3: Grant Matching & AI
- [ ] Build Hybrid Search API (Boolean filters + Vector Similarity).
- [ ] Implement AI Orchestration using LangGraph (Researcher, Writer, Auditor).
- [ ] Setup PDF Assembly service using
pdf-liband@react-pdf/renderer. - [ ] Implement Table of Contents generation with global pagination logic.
Phase 4: Dashboard & Extension
- [ ] Build Kanban Dashboard using
dnd-kitand Radix UI. - [ ] Implement Twilio Message Scheduling for T-14, T-7, and T-1 reminders.
- [ ] Develop Manifest V3 Browser Extension with Chrome Side Panel API.
- [ ] Implement "One-Click Save" in extension using shared session cookies.
Phase 5: Launch & Optimization
- [ ] Perform Post-Quantum resilience audit on vault keys.
- [ ] Run automated WCAG 2.2 accessibility testing.
- [ ] Deploy to Vercel (Frontend) and high-performance Python runtime (AI Backend).
- [ ] Configure Twilio Advanced Opt-Out keywords (
REVOKE,STOP).