# How to Vibe Code Your Own Paperpal (and Stop Paying for It)

> The AI writing tool built for academic research

- Site: https://paperpal.com
- Category: Academic Writing & Research Assistant
- Platforms: Web app, Chrome extension
- Verdict: **Solid side project** (62/100 vibecodeable)
- Estimated effort: 2-3 weeks part-time

## Verdict

You can build a polished personal subset with an AI coding agent, but replicating its massive proprietary academic corpus and journal submission pipeline is out of reach.

Paperpal is an enterprise-backed platform leveraging 24 years of human editing data from Cactus Communications. While you can easily spin up a Next.js web app with the Vercel AI SDK, RAG over open scholarly datasets (via Semantic Scholar or Crossref APIs), and a citation generator, you will quickly hit the limits of generic LLMs when trying to replicate true academic register and consistency checking. Building a personal research workbench for your own papers is entirely viable over 2-3 weekends; replicating Paperpal as a commercial product is a multi-year engineering undertaking.

### What you can't replicate

- 24 years of proprietary human editorial feedback and STM fine-tuning data
- Exclusive direct integration pipelines with 1,500+ top-tier academic journals
- Proprietary index of 250 million verified research articles with enterprise licensing

## What it does

An AI-powered academic writing, research assistant, and editorial platform featuring context-aware grammar correction, academic paraphrasing, citation generation, literature search across 250M+ articles, and submission readiness checks.

### Core features

- Academic grammar and tone proofreading
- Context-aware academic paraphrasing
- Literature search and Q&A across 250M+ articles
- Automated reference and citation generator (10,000+ styles)
- Plagiarism and AI content detection scanner
- Consistency and number/table verification
- Pre-submission journal readiness checks
- Multilingual academic translation preserving terminology

## The business

### Pricing

- Free Plan: Free — Basic allocations for researchers and students
- Prime Plan: $25/mo — Full access for active researchers

Founded 2021.
Team size: 50-200.

## The hard parts

- Fine-tuning or prompting LLMs to maintain strict academic register without flattening technical terminology
- Building low-latency RAG and semantic search over hundreds of millions of scholarly articles
- Developing document-level consistency checking algorithms for numbers, tables, and stats
- Engineering robust document text segmentation and parsing for complex LaTeX and Word formats

## How to vibe code Paperpal

### Prerequisites

- Node.js (free): Required runtime for running Next.js and frontend tooling.
- GitHub (free): Code repository hosting and version control for AI coding workflows.

### Recommended AI tools

- Claude Code: Best-in-class terminal agent for scaffolding the full Next.js stack, wiring RAG pipelines, and writing complex React components.
- Cursor: Ideal for iterative refinement of the editor interface, sidebar panels, and real-time text diff views.

### Stack

- Frontend: Next.js with Tailwind CSS and Shadcn/ui
- Backend: Next.js App Router API Routes
- Database: Turso (libSQL/SQLite at the edge)
- Auth: better-auth
- Payments: none
- Other: Vercel AI SDK, Exa API for scholarly search, Cohere Rerank for retrieval accuracy

### Hosting

- Cloudflare (Hosting the Next.js frontend, API endpoints, and serverless edge functions.): $0-5/mo

### Build guide

1. **Project Scaffolding & Database Setup** — Initialize the Next.js project with Tailwind CSS, Shadcn/ui, and configure Turso with better-auth for personal user management.

```
Create a new Next.js project using the App Router and TypeScript. Install Tailwind CSS and set up Shadcn/ui primitives. Configure Turso (libSQL) as the database provider with Drizzle ORM. Implement better-auth with email/password authentication so a solo user can log in securely. Set up the foundational layout with a responsive sidebar and dark/light mode toggle. Ensure all environment variables are properly structured in a .env.example file. Write clear setup instructions in a CLAUDE.md file in the root directory for future agent iterations.
```

2. **Academic Document Editor & Workspace** — Build a split-pane rich text editor interface where users can write manuscripts, track word counts, and view inline AI suggestions.

```
Build an academic document workspace in Next.js featuring a rich text editor (using TipTap or a similar markdown/HTML editor) on the left and an interactive AI assistant sidebar on the right. The editor must track word counts, section headings (Abstract, Introduction, Methods, Results, Discussion), and support inline comment highlights. The sidebar should have tabs for 'Writing & Improve', 'Search & Cite', and 'Check & Finalize'. Implement local autosave to the Turso database with a status indicator ('Saved', 'Saving...').
```

3. **Academic NLP & Grammar Correction Engine** — Integrate the Vercel AI SDK with Anthropic Claude to analyze text blocks, offering precise academic grammar and tone suggestions without destroying technical terminology.

```
Implement an AI grammar and tone correction endpoint using the Vercel AI SDK and Anthropic Claude. Create a server action that accepts a paragraph of academic text and returns structured JSON containing suggested edits, error categories (grammar, tone, word choice, clarity), and rationale. Ensure the prompt explicitly instructs the model to preserve technical terms, equations, and formal scholarly register without flattening domain vocabulary. Build a React component in the sidebar that displays these suggestions as interactive diff cards with 'Accept' and 'Reject' buttons.
```

4. **Literature Search & Citation Generator** — Integrate Exa API and Crossref/Semantic Scholar metadata to search scholarly literature, summarize findings, and auto-format citations.

```
Build a 'Search & Cite' module in the sidebar. Create an API route that queries the Exa API and scholarly search endpoints for academic papers based on natural language queries. Display the top results with title, authors, publication year, abstract snippet, and citation count. Add a 'Cite in Text' button that inserts a formatted citation tag (e.g., APA or IEEE style) into the active editor cursor position, and maintain a persistent 'Citation Library' table in Turso linked to the current document ID.
```

5. **Consistency & Submission Readiness Checks** — Build heuristic and LLM-powered pre-submission audit tools to scan documents for numerical consistency, spelling uniformity, and structural readiness.

```
Create a 'Check & Finalize' module that performs pre-submission audits on the manuscript text. Implement checks for: (1) number/table label consistency (checking if Table 1 is referenced in text), (2) spelling variant uniformity (US vs UK English), and (3) statistical reporting completeness. Display the results in a submission readiness checklist dashboard with pass/warning/fail badges and actionable fix recommendations generated via structured LLM output.
```

6. **Polish, Error Handling & Deployment** — Add robust error boundaries, loading skeletons, toast notifications, and deploy the application to Cloudflare.

```
Audit the entire application for error states, loading skeletons, and responsive layout bugs on mobile and tablet screens. Add toast notifications (using Sonner) for user actions like saving documents, accepting suggestions, and copying citations. Configure production build scripts and ensure seamless deployment compatibility with Cloudflare Pages/Workers. Write a comprehensive README documenting local development setup, environment variables, and architecture.
```

### Cost vs paying

**Starting costs (one-time):**

- Custom Domain (optional): $12/year
- Exa & LLM API Credits: $10 one-time
- Total: ~$22 one-time

**Ongoing costs (monthly):**

- Cloudflare Workers & D1 Database: $0-5/mo
- LLM & Search API Usage (Anthropic / Exa): $5-15/mo
- Total: ~$10-20/mo

- Paying for the SaaS instead: $25/mo
- Build time: 25-35 hours
- AI tool credits: $20/mo (Claude Pro)
- Break-even: 1 month

## Sources

- [Paperpal Official Website](https://paperpal.com)
- [Paperpal Pricing Page](https://paperpal.com/pricing)
- [Paperpal Data Security & Trust Center](https://paperpal.com/data-security)