The LawDiver MCP server puts 9M+ verified U.S. opinions inside your AI assistant. It searches, reads, and cite-checks against the real corpus — so the authority in the answer is authority that exists.
A general-purpose model asked for a citation will produce something that looks like one. Courts have sanctioned lawyers over exactly that. An MCP connection changes where the authority comes from: instead of recalling a case, the assistant queries a database of real opinions and quotes what it finds — and you can follow every cite back to the full text.
Three tools, the same ones the REST API exposes, called by the model as it works.
Semantic, keyword, boolean, and citation search across federal and state opinions, scoped by jurisdiction and court. The assistant gets the passage that made each case relevant, not just a title.
Pull the full text and structured metadata — court, dates, holdings, citations — so the model reasons over the actual decision instead of a summary of it.
Confirm a citation exists, resolve it to the real case, and report treatment: overruled, questioned, distinguished, or still good law.
Scope a session to one state, one circuit, or the U.S. Supreme Court, and the assistant cannot wander into authority that does not bind your court.
Sign up with email and a password. You get an API key that covers both MCP and the REST API.
Paste the LawDiver server URL and your key into your assistant’s connector settings. No local install, no Docker, nothing to keep running.
The assistant decides when to search, read, or cite-check. You see which tool it called and what came back — the research is auditable, not a black box.
Every result carries the case name, citation, court, and date, and links to the full opinion on LawDiver so you can verify before you rely on it.
The MCP tools wrap the LawDiver API. The developer reference covers authentication, every endpoint and field, error codes, idempotency, and runnable examples.