Verify the claim before you cite it.

CiteVahti tests every manuscript claim against its candidate papers through a blinded human → AI → adjudication workflow, and a reference enters Zotero only as the audited, undoable last step of your decision. Free and local-first.

Get started View on GitHub

[oo] verified [o] needs support [r] review needed [d] decision recorded

How it works — the manuscript as a test suite

Claim detected. Each manuscript claim becomes a test case.
Evidence found. Staged from PubMed — candidates, not citations.
Support, not just topic. Does the paper actually support this claim? (PICO match)
You decide — blinded. The AI's opinion stays hidden until you rate. The human is the decider.
Guarded write. Preview → confirm → write to Zotero, with undo. No silent writes; dedupe fails closed.
Integrity report. Every claim, its state, and a hash-chained audit trail.

Get started

Public beta — local-first, single-user, PubMed-only. Now on PyPI; VS Code Marketplace coming.

# 1. install from PyPI
pip install "citevahti[mcp]"

# 2. set up — connect Zotero with one paste, no hand-crafted keys
citevahti init
citevahti connect-zotero

# 3. start verifying claims
citevahti start

# full guide: github.com/heidihelena/citevahti/blob/main/docs/QUICKSTART.md

Read the Quickstart →

Safe by design

You decide. The AI is a blinded, advisory second rater — never decisive, never silent.
No silent writes. Every Zotero write is preview → confirm → commit, and undoable.
Your key is yours. Stored only in the OS keychain; reads need no key at all.
Local-first & private. Runs on your laptop. No account, no telemetry, no fabricated citations.
Auditable. A hash-chained log records the reasoning — reportable in a methods section.