Academic research plugins for Claude Code and Antigravity: MCP-grounded citations, empirical integrity, systematic reviews, Zotero operations, and parallel-critic manuscript revision.
Inside the Claude Code chat (Desktop or CLI):
/plugin marketplace add mronkko/claude-academic-research
/plugin install academic-research@mronkko
After install, run /setup once to configure API keys, MCP servers, and
permission rules. The wizard is chat-driven — no terminal required.
Inside the Antigravity session/CLI (or terminal):
# Install the main academic-research plugin
agy plugin install "https://github.com/mronkko/claude-academic-research.git"
# (Optional) Install the editorial-tools sub-plugin
agy plugin install "https://github.com/mronkko/claude-academic-research.git/editorial-tools"Works on Windows, macOS, and Linux. Windows users do not need WSL or
Git Bash — native cmd and PowerShell are supported, and CI verifies
every commit against windows-latest.
Nine user-invocable skills:
| Skill | Mode | Purpose |
|---|---|---|
grounded-citations |
rule-book (eager) | Every citation = Zotero BBT key + externalised source consultation; drop claims the source doesn't support. |
empirical-integrity |
rule-book (eager) | Every number in prose must come from an inline expression reading analysis/results/. |
manuscript-revision |
rule-book (eager) | Parallel-critic revision loop is the default revision protocol — delegates to /critic-loop. |
academic-style |
rule-book (eager) | House-style conventions at drafting time — APA citations, voice, tense, hedging, synthesis-over-enumeration, terminology. |
systematic-review |
procedure (explicit) | End-to-end SLR pipeline from search → screening → coding → export. |
zotero-operations |
procedure (explicit) | Import, dedup, enrich, attach PDFs, maintain BBT keys. |
fact-check |
procedure (explicit) | Verify citations and quantitative claims against sources. |
critic-loop |
procedure (explicit) | Run 4 parallel critics (evidence / method / argument / expert) until no MAJOR issues remain. |
setup |
procedure (explicit) | Chat-driven configuration wizard for first-time install. |
Plus one sub-skill, verifying-citations — not invoked directly by
users; loaded by fact-check and critic-loop's evidence critic to
share one citation-verification rule-book between the two callers.
- Plugin scripts run with
uvand use PEP 723 inline dependency declarations — no venv, norequirements.txt, nopip install. - Secrets live in
~/.config/academic-research/config.toml(mode 0600) or environment variables. Environment variables take precedence. - A
permissions.denyrule blocks Claude's Read tool from accessing the config file, so API keys never enter a conversation context. - Reference management goes through Zotero via the Better BibTeX local
JSON-RPC endpoint (
localhost:23119).
.claude-plugin/
plugin.json # plugin manifest
marketplace.json # self-hosted marketplace catalog
skills/ # SKILL.md per skill
scripts/
core/ # config loader/writer, llm provider primitives
sources/ # predatory-journal (Beall's list) checks
pipelines/ # orchestrator scripts (search, enrich_*, abstract_screen,
# fulltext_code, generate_bib, ...) plus:
fetchers/ # per-provider abstract/PDF fetchers (Crossref,
# OpenAlex, ScienceDirect, browser/ for
# Cloudflare-gated publishers)
searchers/ # per-database search backends (Scopus, WoS,
# OpenAlex, Semantic Scholar)
setup/ # first-run configuration wizard + scaffold helpers
editorial-tools/ # second, independently-versioned plugin (peer-reviewer
# suggestion skill) — see Install and Load above
tests/unit/ # pytest + responses mocks
.github/workflows/ci.yml # pytest + ruff on push/PR
MIT. See LICENSE.