DocketXDocketX / In-house counsel
The in-house legal AI that shows its work
Every legal AI vendor is selling your team speed. The thing a general counsel actually answers for is what went OUT under the department's name — and one hallucinated citation in an advice memo costs more credibility than a year of speed buys. DocketX was built backwards from that failure.
What your team gets today
Verification as infrastructure, not a promise
Any model, grounded and checked
Keep the models your team already uses. Before answering, we retrieve the controlling law; after, every citation is verified against 18.1 million real citations — what fails is removed and disclosed, never silently served.
An audit trail per request
For any answer: what was searched, what was checked, what was removed, and why — retrievable by request id. When someone asks "where did this authority come from," the answer is a record, not a reconstruction.
A data policy you can forward to outside counsel
Published, short, and specific: what is stored, what never is, and which upstream providers see what — selectable per key, including bring-your-own-key to your existing model contracts.
Matter files with quarantine
Upload matter documents and query against them; embedded instructions in hostile documents are detected and quarantined, and matter content never enters the shared library.
Zero-hallucination, measured
Across 348 adversarial and real answers on three models, zero unverifiable citations survived to a final answer. The number is measured on the production system, not promised.
Per-query pricing
$0.01 per retrieval query plus model usage at cost. No seats, no annual contract, prepaid credits your finance team will understand in one line.
The honest scope
What we don't do
DocketX is a research and verification layer, not a contract-lifecycle tool: no redlining, no clause libraries, no DMS integration. Coverage today is deepest in Texas (615,391 appellate decisions, 122,681 statute sections), with the U.S. Supreme Court, word-for-word rules for Texas, federal and Washington courts, and a citation-existence table covering every U.S. reporter. Washington case law is being added, more states follow, and the API reports exactly what is served, per state, with counts and dates — so your team never has to guess what we cover.
Start
Two ways in
Self-serve API
Mint a key, $10 free trial allowance, wire it into whatever your team already uses. Docs cover the production checklist.
Get a key →Read the docs first
Grounding, verification, matters, data policy, errors and limits — written for the engineer your team will actually assign.
Docs →Prove the engine in 10 seconds
Paste any citation from any memo into the free checker.
Check a citation →Need the engine inside your perimeter? Private inference runs the same verification on hardware you control.
See also: the AI general counsel overview — how the four gates work and what stays with your lawyers.
Not a law firm; no legal advice. Verification proves a citation exists and its name matches the record — good-law status and support for a proposition remain legal judgment, which is why you exist.