A Claude Code plugin that takes you from a rough topic idea to a polished, peer-reviewed academic paper — with every citation verified against real databases.
A full research pipeline from question design to final paper, powered by specialized AI agents.
Searches Semantic Scholar, OpenAlex, CORE, arXiv, PhilPapers, SEP, IEP, and NDPR. Every citation verified via CrossRef. Open-access PDFs auto-downloaded.
Builds a structured argument skeleton (claim / evidence / counterargument / rebuttal), then expands into full prose with Chicago author-date citations.
Three rounds with two AI personas — Athena (reviewer) and Calliope (reviser). Covers argument validity, evidence quality, and writing polish.
Close reading and logical annotation of primary philosophical texts. Identifies logical connectors, argument structure, and integration points.
Run the entire pipeline with configurable autonomy (cautious / moderate / full). Self-heals when quality gates fail — retries automatically before asking for help.
Three validation checkpoints between phases ensure coverage, citation integrity, and revision completeness. No shortcuts to the finish line.
Each stage builds on the last. Quality gates prevent issues from propagating downstream.
/claudistotle:research-designDefine your research question, scope, and methodology. Output: research-proposal.md
/claudistotle:literature-review6-phase automated review: environment check → domain decomposition → parallel research → synthesis planning → section writing → assembly & export.
/claudistotle:validate Mode AQuality gate: Is literature coverage sufficient? Fails → loop back for supplementary search.
/claudistotle:refineRe-evaluate the research question: proceed, narrow, pivot, or expand.
/claudistotle:draftStep 1: argument skeleton. Step 2: full academic prose with Chicago citations.
/claudistotle:validate Mode BQuality gate: Citation integrity check. Every reference verified against CrossRef and Semantic Scholar.
/claudistotle:peer-reviewThree rounds: argument validity → evidence quality → writing polish. Two AI personas iterate on your draft.
/claudistotle:validate Mode CQuality gate: Change record completeness. Ensures all reviewer feedback has been addressed.
Your reviewed, citation-verified paper is ready.
Works on macOS, Linux, and Windows (Git Bash). Takes about 5 minutes.
You need Claude Code (CLI or Desktop), Python 3.9+, and uv.
# Install uv (fast Python package manager)
curl -LsSf https://astral.sh/uv/install.sh | sh
# Install jq (needed for BibTeX hooks)
brew install jq # macOS
# apt install jq # LinuxAdd from the marketplace, or install locally for development.
# From marketplace
/plugin marketplace add Rlin1027/claudistotle
/plugin install claudistotle@claudistotle
# Or local install
claude plugin install --plugin-dir ./claudistotle-plugin --scope projectClaudistotle needs API access to search academic databases. The setup wizard handles this automatically.
# Run the setup wizard inside Claude Code
/claudistotle:setup
# It creates ${CLAUDE_PLUGIN_DATA}/.env with:
BRAVE_API_KEY=your-key # brave.com/search/api (free tier)
CROSSREF_MAILTO=your@email.com # No signup needed
S2_API_KEY=your-key # semanticscholar.org (free, recommended)
OPENALEX_EMAIL=your@email.com # No signup neededJust tell Claude what you want to research. The plugin handles the rest.
# In Claude Code, simply type:
I need a literature review on the extended mind thesis
and its implications for cognitive offloading.
# Claude invokes /claudistotle:literature-review
# Output lands in reviews/[your-topic]/Each command works independently. Use the full pipeline or just the parts you need.
| Command | Description | Phase |
|---|---|---|
/claudistotle:setup | First-time configuration wizard | Pre |
/claudistotle:help | Interactive usage guide and tutorial | Pre |
/claudistotle:research-design | Define research question, scope, and methodology | 1 |
/claudistotle:literature-review | Automated 6-phase literature review (search, verify, synthesize) | 2 |
/claudistotle:text-commentary | Primary text close reading and logical annotation | Any |
/claudistotle:validate | Quality gates between pipeline phases (Modes A, B, C) | Gates |
/claudistotle:refine | Research question refinement based on findings | 3 |
/claudistotle:draft | Two-step academic writing (skeleton + prose) | 4 |
/claudistotle:peer-review | Three-round peer review simulation | 5 |
/claudistotle:feedback | External advisor feedback integration and routing | Any |
/claudistotle:autopilot | Run the entire pipeline end-to-end | All |
/claudistotle:philosophy-research | Direct academic database search | Any |
Everything is saved to reviews/[project-name]/ in your repository.
Mention you want a "literature review" explicitly. Otherwise Claude may try to help without invoking the full orchestrated pipeline.
Type /model in Claude Code and select Sonnet. Reserve Opus for complex synthesis. This saves significant API costs.
If a session is interrupted, just say "continue". Claude reads PROGRESS.md and picks up where it left off.
The .bib files work with Zotero, Mendeley, BibDesk, and LaTeX/pandoc. Open in a text editor for rich metadata.
You don't need the full pipeline. Run /claudistotle:literature-review alone, or jump to /claudistotle:draft if you have your research.
Claudistotle searches open databases. For paywalled articles, download PDFs manually to sources/secondary/.