The AI-native platform that turns raw RFPs into winning proposals — automatically.

Companies juggle dozens of RFPs every quarter, each demanding hours of cross-referencing inventory, benchmarking competitors, and writing polished proposals from scratch. BidForge eliminates that grind. Upload an RFP document and a five-agent AI pipeline takes over — parsing requirements, matching line items against your product catalog via RAG, pulling live market pricing through SerpAPI, and synthesizing a competitive pricing strategy. Before the final PDF is generated, you review and select from recommended pricing options, keeping humans in the loop where it matters. The result is a branded, ready-to-send proposal document uploaded to S3 in minutes, not days.

Homepage	Dashboard
Inventory Management	RFP Exploration
Summarised Pricing Options	Final Proposal Artifact

• 5-Agent AI Pipeline — Parse → Explore → Summarise → Generate, each agent is independent and composable
• Intelligent RFP Parsing — Extracts requirements, flags missing fields, and cleans raw text using Gemini 2.5 Flash via OpenRouter
• Inventory RAG — Matches RFP line items against your uploaded product catalog (PDF, CSV, TXT) stored in S3
• Competitor Intelligence — Analyzes uploaded competitor files + fetches live market prices via SerpAPI (IQR-cleaned median pricing)
• Semantic Search on Past RFPs — Vector embeddings (HuggingFace all-MiniLM-L6-v2) stored in MongoDB Atlas with $vectorSearch for cross-RFP learning
• Human-in-the-Loop Pricing — Summariser presents 2–3 pricing strategy options per line item; user selects before document generation
• PDF Export — Final proposal rendered from Markdown → PDF via md-to-pdf (Puppeteer) with branded styling, uploaded to S3
• Email Ingestion — Cloudflare Worker receives inbound emails and triggers the pipeline automatically
• Company Auth — Email verification flow via Resend, JWT-based session middleware
• Monorepo — Bun workspaces with shared common package for type-safe event schemas

sequenceDiagram
    participant U as User / Email
    participant FE as Frontend (React 19)
    participant BE as Backend (Fastify)
    participant P as Parser Agent
    participant E as Explore Agents
    participant S as Summariser Agent
    participant D as Document Agent
    participant DB as MongoDB Atlas
    participant S3 as AWS S3

    U->>FE: Upload RFP document
    FE->>BE: POST /api/rfp/upload
    BE->>DB: Store RFP + embedding
    FE->>BE: POST /api/rfp/:id/parse
    BE->>P: runParseTextStep()
    P-->>BE: Parsed requirements + missing fields
    BE->>DB: Update status → "parsed"
    FE->>BE: POST /api/rfp/:id/explore
    BE->>E: runExploreStep() [Inventory + Competitor in parallel]
    E-->>BE: Inventory analysis + Competitor pricing markdown
    BE->>DB: Update status → "explored"
    FE->>BE: POST /api/rfp/:id/summarise
    BE->>S: runSummariseStep() + search_past_rfps tool
    S-->>BE: Pricing options per line item
    FE->>U: Show pricing options for review
    U->>FE: Select preferred pricing options
    FE->>BE: POST /api/rfp/:id/generate-document
    BE->>D: generateFinalDocument()
    D-->>BE: Markdown proposal
    BE->>S3: Upload PDF
    BE-->>FE: Return S3 document URL

bidforge/
├── apps/
│   ├── backend/          # Fastify API + AI agents (Bun)
│   │   ├── src/
│   │   │   ├── agent/    # 5 AI agents + LLM config
│   │   │   ├── routes/   # auth/, company/, rfp/
│   │   │   └── utils/    # embedding, pricing, s3-client
│   │   └── prisma/       # MongoDB schema
│   ├── frontend/         # React 19 + TanStack Router
│   │   └── src/
│   │       ├── components/
│   │       ├── routes/
│   │       └── store/
│   ├── email-worker/     # Cloudflare Worker (email ingestion)
│   └── common/           # Shared Zod schemas (EmailEvent)
└── assets/               # Screenshots & banner

• Bun ≥ 1.3
• MongoDB Atlas cluster with a Vector Search index named vector_index on the RFP collection, path embedding
• AWS S3 bucket
• Accounts / API keys for: OpenRouter, HuggingFace, SerpAPI, Resend

git clone https://github.com/yashraj-n/bidforge.git
cd bidforge
bun install

Create a file at apps/backend/.env with the following variables:

cd apps/backend
bunx prisma generate
bunx prisma db push

# Backend (port 9000)
cd apps/backend && bun run dev

# Frontend (port 3000)
cd apps/frontend && bun run dev

stateDiagram-v2
    [*] --> uploaded
    uploaded --> parsed : /parse
    parsed --> exploring : /explore triggered
    exploring --> explored : agents complete
    explored --> summarised : /summarise
    summarised --> generating_document : /generate-document
    generating_document --> completed : PDF uploaded to S3
    uploaded --> rejected : /reject
    parsed --> rejected : /reject
    explored --> rejected : /reject
    exploring --> failed : agent error
    generating_document --> failed : PDF error

Contributions are welcome! Here's how to get started:

1. Fork the repository
2. Create a feature branch: git checkout -b feat/your-feature
3. Commit your changes: use Conventional Commits (feat:, fix:, chore:, docs:)
4. Push to your fork and open a Pull Request

Branch naming conventions:

This project is licensed under the Apache License 2.0 — see the LICENSE file for details.

If BidForge saves your team hours on proposals, consider giving it a ⭐
Built with ❤️ by @yashraj-n, @rsk807, @dhiraj-rajput, @ashish9925