OpenDraft is a free, open-source Python engine designed to generate academic research paper drafts. It addresses the common issue of AI-generated content hallucinating citations by verifying every source against academic databases like CrossRef, OpenAlex, Semantic Scholar, and arXiv. Targeting researchers, academics, and graduate students, it offers a significant benefit by producing long-form documents (20,000+ words) with verifiable citations, alongside multiple export options, and a free hosted version for ease of use.

How It Works

OpenDraft employs a multi-agent architecture, utilizing 19 specialized AI agents that mimic a research team. These agents handle distinct phases: research paper discovery, structural outlining, content drafting, citation verification, language polishing, and final export. This modular approach allows for focused execution on each task, ensuring that citations are rigorously checked against real-world academic literature databases, a key differentiator from general-purpose LLMs that often fabricate sources.

Quick Start & Requirements

• Primary install: Clone the repository and run pip install -r requirements.txt.
• Prerequisites: Python 3.10+, a supported AI model API key (e.g., Google Gemini, OpenAI, Anthropic). For PDF/audio features, additional dependencies and API keys (ElevenLabs) are required.
• Setup time: Approximately 10 minutes for local installation.
• Links: Hosted version: OpenPaper.dev, Contributing guidelines: CONTRIBUTING.md.

Highlighted Details

• Verified Citations: Every citation is cross-referenced with CrossRef, OpenAlex, Semantic Scholar, and arXiv to ensure it links to a real publication.
• Long-Form Content: Capable of generating drafts exceeding 20,000 words, suitable for theses and extensive literature reviews.
• Multi-Format Export: Supports export to PDF, Microsoft Word (.docx), and LaTeX formats.
• Extensive Language Support: Drafts can be generated in over 57 languages.
• Standalone Tools: Includes TL;DR (5-bullet summary) and Audio Digest (60-second audio briefing) for quick analysis of any research paper.
• Data Fetching & Revision: Integrates with statistical APIs (World Bank, Eurostat, Our World in Data) and offers AI-assisted draft revision capabilities.

Maintenance & Community

The project is actively maintained, with contributions welcomed via GitHub Discussions and Issues. The repository includes specific documentation for maintainers regarding its internal workflows, including automated PR reviews and regression test generation using AI models.

Licensing & Compatibility

OpenDraft is released under the MIT License, which permits commercial use, modification, and distribution without significant restrictions, making it highly compatible for integration into commercial products or services.

Limitations & Caveats

OpenDraft is explicitly not designed for generating final papers without human oversight, cheating, or replacing human researchers. Users are advised to review, edit, and add their own analysis to all generated drafts, and to verify their institution's policy on AI-assisted work.