⚖️ LEGAL

How to check legal citations and research state law with Jev

Two legal tools built on Jev in the same week. LegalJev checks that the cases a document cites actually say what it claims, and it’s free. Monid answers one legal question across every state and quotes the statutes. Here’s how to use both, and the pattern behind them.

Written September 28, 2026 · By Ben Broch
statute sections read
134
states with the same rule
42
for the whole run
80¢

From @Jasperli0122’s public demo

The tools: one free app, one paid API

LegalJev is a free web app by Jozef, the maker of StealthGPT. There’s no code published, but there’s nothing to install either: no signup, and he says it doesn’t store your data.

Monid is a paid API by Jasper Li’s team that gives an AI agent access to thousands of tools through one key. Its connector code is open source on GitHub (MIT), but the legal demo itself isn’t published as code. You set it up by giving your agent one instruction.

Both are beta research tools, not legal advice. Check every result against the source before you rely on it.

01

LegalJev: catch the fake case before the judge does

AI writing tools will cite cases that don’t exist, or cases that hold the opposite of what they’re cited for. In the demo, LegalJev checked every citation and claim in a 7-page memo in under 9 seconds. It found one potentially fabricated citation, plus several claims that the cited cases don’t support or directly contradict.

To use it, open legaljev.com, choose Audit citations, and paste your text or add up to 20 PDF, DOCX, TXT or Markdown files. A second mode, Find authorities, looks for case law that supports an argument.

02

Monid: one question, every state

The question in the demo was “how long does a landlord have to return a security deposit?” It read 134 statute sections across 45 states, found the same rule in 42, and quoted the exact deadline from 12 statutes. The header on screen reads 17 calls, $0.80, 3.6 seconds.

The law itself comes from Vaquill, a database of U.S. federal and state statutes that Monid sells access to per call. Jev does the judging on top. Setting it up is one instruction to your agent:

Paste at your agent (from Monid’s page)
# From monid.ai/blog/reading-law. Paste at your agent (I haven't run this myself)
set up https://monid.ai/SKILL.md, then use vaquill on monid to answer <your legal question>: cite the section, quote the exact subsection, and link the official source

Pricing and the full walkthrough: monid.ai/blog/reading-law. Monid is paid per call from a prepaid balance.

The same pattern with Jev

Both tools come down to the same move: put a claim next to its source and ask whether they agree. One call per claim:

A citation that can’t be found at all is the loudest flag, and it doesn’t need a model. Jev earns its place on the subtler problem: a real case that’s cited for something it doesn’t say. It’s the same shape as the prospecting guide’s message check, with a legal claim in place of a message.

New to Jev? Start with the setup guide: key, first call, working response. Then come back.

Let an agent build your own checker

If you want this inside your own workflow instead of a website, paste this at Claude Code or Codex. It only flags problems and never rewrites anything. The 0.50 bars are starting points I haven’t run yet.

Paste at Claude Code / Codex
Build me a citation checker for a legal document using Jev through the Vercel AI Gateway. It flags problems for a lawyer to review; it never edits the document.

1. Ask me for the document (a brief or memo, as PDF or text). Pull out every sentence that cites a case, and the case it cites.
2. For each citation, fetch the cited opinion from a free public source (CourtListener has an API) and keep the passages most related to the claim. If the case can't be found at all, flag it NOT FOUND and skip step 3 for it.
3. For each claim make ONE call: experimental_evaluate from the ai package, model typesafe-ai/jev. The state is "CLAIM: <the sentence from my document> / OPINION EXCERPTS: <the passages>". Ask these boolean questions together in that one call:
   - supports: "The opinion excerpts support the claim as written."
   - contradicts: "The opinion excerpts say the opposite of the claim."
   - quote: "Any quotation in the claim appears in the excerpts word for word."
4. Flag a claim if supports is under 0.50, contradicts is over 0.50, or quote is under 0.50. Sort NOT FOUND first, then contradictions, then weak support.
5. Run 3 calls at a time; bursts get rate-limited. Write citations.csv (claim, case, verdict, scores, the excerpt it relied on).

Show me the flags with the excerpt next to each. This is a first pass for a lawyer to check, not legal advice.

The honest part

I haven’t run either tool on a real case. The numbers are from @jozef_gherman’s and @Jasperli0122’s posts, both from September 22, 2026. The Monid run is Jev plus Vaquill’s statute database through Monid, not Jev alone.

None of this is legal advice, and both tools say so. They’re fast first passes. A lawyer still reads the case before anyone relies on it.

Don’t have Jev running yet?

The setup guide is one key and a few lines of code, and you’ll have a working response in a few minutes.

Get started with Jev →

More field guides

All guides →