An AI Co-Author for Grant Season Survivors

What it does

It’s an orchestration layer that turns scattered datasets and rough ideas into formatted scientific documents—IMRaD papers, NSF/NIH grant proposals, LaTeX conference posters, clinical case reports, and literature reviews. The tool feeds your prompts and uploaded files to Claude, optionally queries Perplexity Sonar Pro for live literature search via OpenRouter, and spits out structured outputs with auto-generated BibTeX and inline figures. You can drive it through a Claude Code plugin, an async Python API, or a standalone CLI.

The interesting bit

The project treats scientific writing as a pipeline rather than a chat session: it auto-sorts dropped files into figures/ and data/, runs a quantitative peer-review pass with an 8-dimension ScholarEval score, and can summon AI-generated schematics through Nano Banana Pro. That makes it less a chatbot wrapper and more an opinionated workflow engine for academics who already have results but dread the formatting.

Key highlights

• Generates domain-specific documents (Nature/Science-style papers, DARPA grants, clinical reports) with structured sections
• Performs real-time literature lookup through Perplexity Sonar Pro Search to source citations
• Includes a peer-review skill that scores drafts across eight dimensions before you submit anywhere
• Offers 19+ specialized skills when running as a Claude Code plugin, from hypothesis generation to LaTeX poster layout
• Async Python API with streaming progress updates for programmatic use

Caveats

• The README itself contains typos (“scientiic”, “scientfic”) in its own opening banner.
• The documentation states the tool “works best as a Claude Code plugin,” treating the CLI and Python API as secondary interfaces.
• It promises “publication-ready” documents and “verified citations” without showing sample outputs or accuracy benchmarks in the repository.

Verdict

Worth a look if you are a researcher sitting on a pile of CSVs and microscopy images who would rather automate the LaTeX wrangling than learn it. Skip it if you were hoping for an open, offline scientific writing stack—this is a cloud-API orchestration tool with keys attached.