Upload your contract. Get your invoice.
Invox is a free, AI-powered invoice tool for freelancers, agencies, and small businesses. Upload a project contract or document, describe your work in plain English, and get a polished, ready-to-send invoice in seconds — no forms, no templates to fill out manually, no account required.
"I used to spend 30+ minutes making invoices in Adobe Illustrator. They still looked mediocre. Now I upload my contract, tell the AI my rate, and it's done."
Most invoice tools make you fill out forms. Invox lets you just talk.
Upload a PDF contract and say:
"I completed this project in 14 hours at $40/hour, due in 15 days."
The AI reads your document, extracts the relevant details, builds the invoice, and streams changes back in real time. You can keep refining it conversationally, tweak it in the settings panel, or click any field to edit inline.
Drop in a PDF contract, a project brief, or even a photo of a document. The AI reads it, understands the context, and uses it to build your invoice — no copy-pasting required.
Describe changes in plain English across multiple turns:
"Add a 10% early payment discount"
"Change the due date to end of month"
"Make the template more minimal and switch to blue"
The AI remembers context across messages. No need to repeat yourself.
Standard, Modern, Minimal, Classic, Elegant, Bold, Corporate, Creative, Startup, Receipt, Gradient, and Retro. Each template is fully implemented with its own layout, and supports complete color, font, and spacing customization.
- AI Chat — Describe changes conversationally. The agent handles the rest.
- Settings Panel — Tabbed sidebar for branding, typography, financials, and layout.
- Inline Editing — Click any field on the invoice to edit it directly.
Click the mic icon in the chat input to dictate your message instead of typing. Speech is transcribed server-side using OpenAI's gpt-4o-mini-transcribe model and inserted directly into the input field. Recording caps at 30 seconds automatically. Chat messages have a 1,000-character limit.
One-click export at print-ready resolution (816×1056px). Fonts, colors, and layout are preserved exactly as they appear on screen.
Add a text signature or draw one by hand using a pressure-sensitive canvas.
All invoice data lives in your browser's localStorage. No account, no server-side storage, no data leaving your machine (except AI requests for processing).
Auth and rate limiting are disabled by default — not needed for local use or self-hosted instances. For public deployments, they can be enabled via environment variables to protect your OpenAI key from abuse. See Deploying Publicly below.
Prerequisites: Node.js 18+, an OpenAI API key
git clone https://github.com/aasherkamal216/Invox.git
cd Invox
npm installCreate a .env.local file:
OPENAI_API_KEY=sk-your-key-hereStart the dev server:
npm run devOpen http://localhost:3000.
When hosting Invox for others (e.g. on Vercel), you'll want to protect your OpenAI key from unlimited use. Auth and rate limiting are opt-in — disabled by default, activated by environment variables.
Set these in your Vercel dashboard (leave unset for local/self-hosted):
# Enable auth gate on the AI endpoint
NEXT_PUBLIC_ENABLE_AUTH=true
# Clerk — https://clerk.com (free tier is sufficient)
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=pk_live_...
CLERK_SECRET_KEY=sk_live_...
# Upstash Redis — https://upstash.com (for rate limiting, free tier is sufficient)
UPSTASH_REDIS_REST_URL=https://...upstash.io
UPSTASH_REDIS_REST_TOKEN=...When NEXT_PUBLIC_ENABLE_AUTH=true:
- Users must sign in via Clerk before using the AI assistant
- AI requests are rate-limited to 20 per user per hour via Upstash Redis
- The invoice editor (manual editing, templates, PDF export) remains completely free and unrestricted
When these variables are not set, the app runs as a clean, no-auth tool — exactly as cloned.
npm run dev # Start dev server (hot reload)
npm run build # Production build
npm run start # Run production server
npm run lint # ESLint| Layer | Technology |
|---|---|
| Framework | Next.js 16 (App Router) |
| Language | TypeScript 5 |
| Styling | Tailwind CSS v4 |
| Components | Coss UI (Base UI + Tailwind) |
| AI | OpenAI Agents SDK (@openai/agents), Transcriptions API (gpt-4o-mini-transcribe) |
| PDF Export | jsPDF + html-to-image |
| Auth (optional) | Clerk (@clerk/nextjs) |
| Rate Limiting (optional) | Upstash Ratelimit + Redis |
The AI agent runs entirely server-side via a Next.js Route Handler. The client sends a message and the current invoice state; the server responds with a Server-Sent Events stream of text deltas and invoice patches.
- Invoices are stored locally only — no cross-device sync
- No cross-user collaboration or multi-user support
- The AI agent cannot upload logos, draw signatures, or customize label strings (these are UI-only actions)
Issues and pull requests are welcome. For architectural notes and development conventions, see CLAUDE.md.
- Fork the repository
- Create a feature branch:
git checkout -b feat/my-feature - Commit your changes
- Open a pull request
MIT — see LICENSE for details.