DocketXDocketX / AI general counsel
An AI general counsel that cannot bluff
Every legal AI on the market writes confidently. That is the problem: fluency is not authority, and the failure mode is a memo that reads perfectly and cites a case that does not exist. DocketX is built the other way around — nothing reaches you until its citations have been looked up in the public record, and what could not be verified is removed and named.
The job, honestly divided
What it does, and what stays yours
It does the research function
Answers grounded in real case law, statutes and word-for-word court rules; checks the authority in anything anyone hands you; keeps the record of what was searched and verified.
You keep the judgment
Advice, privilege, strategy and accountability belong to licensed counsel. We are not a law firm and give no legal advice — we make the humans faster and much harder to mislead.
Nobody has to trust us
Every claim on this page is checkable: paste any citation into the free checker, or read the coverage API, which reports per state exactly what we hold and when it was taken.
Under the hood
Four gates, in order, on every answer
1. Ground
Your question retrieves the controlling authority — 615,391 Texas appellate decisions, 122,681 statute sections, 45,660 Supreme Court opinion documents, 6,093 rule sections held verbatim — before the model writes a word.
2. Answer
The model you chose reasons over what was actually retrieved. Keep your own models, ours, or private ones on your hardware.
3. Verify
Every citation is looked up in 18,123,788 citations across 989 reporters. Case names are checked against the courts' own records. Quotations from rules are checked word for word.
4. Disclose
What fails is removed from the answer and the answer says so. Silence about a removal would be its own dishonesty.
Streaming, still gated
Even when text streams live, a citation is held until its verdict arrives — never shown then retracted.
Measured, not promised
348 adversarial and real answers, three models, prompts engineered to bait invented authority: zero unverifiable citations survived.
Deployment
Three shapes, same gate
API, self-serve
A key, a $10 trial allowance, and an endpoint your engineer wires into whatever you already run. $0.01 per retrieval query, model usage at cost.
Get a key →Virtual Counsel Desk
On-demand verified research for teams without an engineer to spare.
The desk →Private inference
The whole engine on hardware you control, when matter content cannot leave your perimeter.
Private →The honest scope
What we will not claim
No contract redlining, no clause library, no document management, no outcome prediction. Coverage is deepest in Texas, with Washington building and other states following on a pipeline that reports its own status. Verification proves a citation exists and its name matches the record — it does not prove the case is still good law or that it supports the proposition, which is exactly the judgment we said stays with your lawyers.
Questions
What buyers ask
What is an AI general counsel?
A system that does what a general counsel's research function does — answer legal questions with real authority, check what others assert, and keep a record — without pretending to be a lawyer or to give legal advice. DocketX is the verified-research layer of that job: grounded answers, every citation checked, an audit trail per request.
Can an AI replace our general counsel?
No, and any vendor telling you otherwise is selling risk. Judgment, privilege, advice, and accountability are a licensed lawyer's. What AI genuinely replaces is the hours of unverified searching around that judgment — and the danger is that most legal AI produces fluent, confident text with citations nobody checked.
How is DocketX different from other legal AI?
Verification is the product, not a feature. Every citation in an answer is looked up in 18,123,788 real citations before you see it; anything unverifiable is removed and disclosed. Across 348 adversarial test answers on three models, zero unverifiable citations survived.
Can it run on our own infrastructure?
Yes — private-pod keys restrict a key to in-house models on hardware you control, with the same grounding and the same citation gate, plus bring-your-own-key support for your existing model contracts.
What does it cost?
$0.01 per retrieval query plus model usage at cost, on prepaid credits. No seats, no annual minimum, and a $10 trial allowance to prove it before you spend.
Talk to us about a desk, private inference, or a state you need next.
DocketX is not a law firm, does not give legal advice, and does not create an attorney-client relationship. It verifies research so licensed humans decide on real authority.