Denario is a modular multi-agent system designed to assist with scientific research, automating tasks from data analysis to paper generation. It targets researchers and power users seeking to streamline their scientific workflows. The system leverages advanced AI agent frameworks to accelerate discovery and publication.

How It Works

Denario employs a multi-agent architecture built upon AG2 and LangGraph, utilizing cmbagent as its research analysis backend. This modular design allows for flexible integration of various AI capabilities. The system orchestrates agents to interpret experimental data, generate research hypotheses, formulate methodologies, perform computations, and ultimately produce scientific papers, offering a novel end-to-end approach to AI-driven scientific discovery.

Quick Start & Requirements

• Primary Install: pip install "denario[app]" (Python 3.12 recommended). Alternatively, use uv init and uv add denario[app].
• Run GUI: denario run
• Prerequisites: Requires LLM API keys for OpenAI, Claude, or Gemini. Docker is available for an all-inclusive environment.
• Links: Project page, Paper, Documentation, Denario GUI repository, Demo web app, YouTube channel.

Highlighted Details

• Denario and its backend Cmbagent were featured in the LSST DESC AI Roadmap (January 2026).
• Cmbagent received first place at the NeurIPS 2025 Fair Universe Competition (December 2025).
• Version 1.0 was released alongside the project's paper on arXiv (November 2025).
• A research paper fully generated by Denario was accepted for publication at the Open Conference of AI Agents for Science 2025.

Maintenance & Community

Contributions are welcomed via pull requests, and issues can be opened for bugs, comments, questions, and suggestions. Direct inquiries can be sent to denario.astropilot.ai@gmail.com.

Licensing & Compatibility

Denario is licensed under the GNU GENERAL PUBLIC LICENSE (GPLv3). This is a strong copyleft license, meaning any derivative works must also be distributed under the GPLv3. This may impose restrictions on integration with proprietary or closed-source software.

Limitations & Caveats

The primary caveat for adoption is the GPLv3 license, which necessitates open-sourcing any modifications or integrations. No other specific limitations or alpha status were mentioned in the provided README.