A Python library +skills, prompts and more artifacts that gives AI agents a reliable, deterministic interface to search, fetch, store, and manage academic references. Think Zotero rebuilt as a machine-native primitive: no GUI, no human workflows — clean functions an agent can call with confidence.
Two paths, one binary. Choose based on how you work with Claude.
For researchers using Claude.ai Projects — no terminal needed. Your agent handles setup.
Download the CLI and the manager skill
Upload both files to your Co-Work session and install the skill.
Ask the agent:
"Set up scholartools in my collection folder"
For technical users — manual setup from your collection directory.
Download the CLI
scht CLIUnzip and make executable
unzip scht-*.zip && chmod +x scht
Run from your collection folder
./scht refs list
Creates .scholartools/config.json on first run.
Research workflow
scht discover "transformer attention"
scht fetch 10.48550/arXiv.1706.03762
scht staging list
scht staging merge
scht files attach vaswani2017 paper.pdf
scht sync push-changelog
Envisioned — not built yet
We envision a transition from isolated reference metadata to shared epistemic objects that enable deeper conversations.