use-citations
github.com/sshah03/use-citations- Category
- AI Agents
- Rank
- No. 2235Tools index
Previous survey · No. 2107 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Research & Education · Data, Retrieval & Knowledge
- Interfaces
- Agent Skill / Plugin
- GitHub
- 1 stars
- Date
About
use-citations is a Claude Code skill that turns research questions in fields like law, tax, and medicine into citation-backed reports: it has Claude save source documents, ground its answers in them, and build an artifact where each sourced phrase opens a slide-out drawer highlighting the exact passage in the original document. It is meant for people with a Claude or GPT subscription who want some of the grounding that domain-specific tools like Harvey, Legora, and OpenEvidence provide, without those tools.
At a glance
use-citations gives Claude Code a workflow for answering from saved sources and checking whether quoted passages appear in them. It is worth trying when you need to trace an answer back to exact text.
Good fit
Researchers and developers working with documents they are allowed to expose to their model.
Consider something else when
Use a local search workflow if the passages cannot be read by Claude or the source must stay outside a model session.
Start here
Ask a question about a short synthetic document. Check the quoted passage and then read its surrounding text to see whether it supports the claim.
Cost and requirements
MIT-licensed. The project documents Claude Code with Python 3.10 or later and a separate Cowork setup.
Limits to know
Finding the quoted text does not establish that the interpretation is sound. Claude can read passages, local copies are retained, and publishing a report uploads it to Claude according to the privacy notes.
Sources checked 2026-11-05. Source-based assessment.
Why it made the leaderboard
Adds citation grounding and traceability to Claude Code research sessions, letting users verify claims against the actual source text instead of trusting an unlinked summary.
Tags
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.