Legal AI infrastructure & MCP tools

Citation-verified legal AI for software and agents.

Add citation verification, quote checking, brief certification, and a drop-in MCP server to your legal software and AI agents — every answer carries its source, verification, and audit trail.

Get an API key Add to Claude in 60 seconds

verdict.io · trust envelopeexample · real response

Request

POST /v1/citations/verify
Authorization: Bearer vd_key_…

{ "text": "Brown v. Board of Education, 347 U.S. 483 (1954)" }

Response envelope

Brown v. Board of Education, 347 U.S. 483 (1954)

verified_exists

source: verdict:cap:11301409 · decided 1954-05-17
evidence: CAP index us/347 CasesMetadata
verify: verdict_volume_range · verdict_cap_index
status: verified
meter: 3 credits

Built on legal data you can cite · works with any MCP client

ClaudeChatGPTCursorVS CodeLangChainCustom agents

10M+ U.S. cases resolvable by citation or id65M+ citation-graph edges behind treatment signalseyecite citation parsingFull audit & verification logs

First products

A legal tool catalog, exposed as APIs and MCP tools.

Call each capability over REST, or mount the whole catalog into your agent through one MCP server. Same envelope, same audit trail, every time.

⚖Live

Brief Verification API

Verify every citation and quotation in a filing at once. Merged flags plus a court-ready certification block for AI-disclosure orders.

POST /v1/briefs/verify

❝Live

Quote Verification API

The 2026 sanctions pattern is fabricated quotes on real cases. Verify quoted language word-for-word against the cited opinion — and when the official scan itself is damaged, Verdict refuses to rule rather than accuse.

POST /v1/quotes/verify

✓Live

Citation Verification API

Parse with eyecite, resolve against the Verdict corpus. Typed labels — verified, not-found, out-of-coverage — each with its evidence.

POST /v1/citations/verify

⚑Beta

Treatment Signals

Is it still cited as good law — and what did the court do with it? Negative-treatment and cautionary language quoting the later opinions, tiered by independent corroboration. Absence of a signal is never clearance.

POST /v1/citations/verify · GET /v1/cases/{id}

⌕Live

Legal Search API

Case-law and statute search with jurisdiction and date filters. Batch retrieval takes up to 500 records per call, billed per record found.

GET /v1/search · POST /v1/cases/batch

MLive

Verdict MCP Server

A drop-in MCP gateway — 13 verified tools for Claude, ChatGPT, Cursor, and your own agents, with published machine-readable consumer guidance.

https://mcp.verdict.io/mcp

§Beta

Case-Law Analysis API

What a case is cited for: holding-level parentheticals, dispositions, and authorship — every signal measured against a blind review panel before it serves.

GET /v1/cases/{id}

◆Live

AI-Output Check

A brief engine tuned for AI drafts, adding typed hallucination rollups. Cheaper than a certification so you can iterate on drafts freely.

POST /v1/ai/verify

≡Live

Verification Log API

Retrieve the audit record behind any prior response — source, verification, review decision, and timestamp.

GET /v1/logs/{request_id}

¶Coming soon

Document Parsing API

Extract parties, dates, and structure from filings and contracts.

POST /v1/documents/parse

◷Coming soon

Deadline Extraction API

Surface deadlines and triggers from documents and dockets.

POST /v1/deadlines

⤓Coming soon

Intake / Triage API

Classify issue type, jurisdiction, and urgency into attorney-ready packets.

POST /v1/intake

Pricing

Pay per use. No seats, no sales calls.

Credits are 1¢ each. You pay for verified work, not licenses — and we charge less when our coverage gaps, because we don't profit from our own gaps.

$9.95

per brief certified

+3¢ per citation over 150 · court-ready certification block

$4.95

per AI-output check

cheaper than the certification — iterate on drafts freely

7¢

per quote checked

word-for-word against the cited opinion

3¢

per citation checked

1¢ on honest coverage-gap outcomes

Under $10 a brief. Subscription tools in this category run $150–300 per user per month.

Free

500 credits every month

≈ $5 of verification monthly — 25 citation checks plus quote credits. No card required.

Start free

Credit packs: $5 · $20 · $100 · $250 (27,500 credits — 10% bonus). Estimates are pre-authorized and the difference refunded; engine failures refund fully; retries never double-charge.

See full pricing

The Verdict MCP gateway

One gateway between your agents and verified legal data.

Connect any MCP client. Verdict handles auth, tenancy, and policy, then routes calls to verified tools and grounded legal sources.

MCP clients

Claude / Desktop

ChatGPT

Cursor / IDE

LangChain agent

Your application

Vverdict.io gateway

● API key & tenant auth

● policy & approval gates

● source + verification envelope

● audit log on every call

Tools & sources

citations.verify

quotes.verify

briefs.verify

search.query

Verdict corpus · 10M+ cases

60-second setup

{
  "mcpServers": {
    "verdict": { "url": "https://mcp.verdict.io/mcp" }
  }
}

Machine-readable consumer guidance is published at /.well-known/verdict-consumer-guidance.json — versioned, and true by construction.

Citation & quote verification

Every citation checked against the real record.

Verdict parses citations with eyecite and resolves them against its own index of the public case-law record. Verified cites pass through; mismatched pincites, unfound authorities, and out-of-coverage cites are each labeled distinctly — never silently trusted, and a coverage gap is never called a fabrication. The same honesty runs the quote lane: a match must survive word-for-word verification, and when the official scan itself is damaged, Verdict refuses to rule rather than accuse.

Research memo · draft

The court applied the qualified-immunity standard from ✓ Pearson v. Callahan, 555 U.S. 223, reaffirming the analysis first set out in ✓ Harlow v. Fitzgerald, 457 U.S. 800. Plaintiff's reliance on ! Doe v. Reed, 561 U.S. 902 is misplaced.

✓

Pearson v. Callahan

555 U.S. 223 · resolved

verified

✓

Harlow v. Fitzgerald

457 U.S. 800 · resolved

verified

Doe v. Reed

pincite not found · review

flagged

⚑

Frye v. United States

negative treatment · beta

treatment

eyecite + public court records · every call logged

Security, tenancy & policy

Built for lawyer supervision, not autonomy.

Source-grounded by default, with human approval before anything is filed, sent, or signed.

🔒

Per-tenant isolation

Every request runs in an isolated tenant context with scoped, revocable keys.

✋

Human approval gates

Built for lawyer supervision, not autonomy — nothing is filed, sent, or signed without review.

≡

Full audit trail

Source, verification, reviewer, and timestamp recorded on every call, retrievable by request id.

↺

Idempotent billing

Retries never double-charge. Replays are marked in the meter, and engine failures refund fully.

⛂

Keys, not passwords

API keys are shown once and stored hashed. Key management requires the OAuth session, so a leaked key cannot mint another.

◎

Coverage disclosed

Every response states what was searched and how complete it was. A coverage gap is never called a fabrication.

The Verdict ecosystem

verdict.io is the infrastructure under it all.

verdict.io

Infrastructure · you are here

APIs, MCP tools, citation verification, and legal data.

verdict.com

Public legal AI

Case-law discovery and the consumer–attorney surface.

verdict.tech

Firm services

Custom legal AI automation and implementation.

verdict.one

Coming

An AI-first operating system for small firms.

Put citation-verified legal AI into production.

Get an API key, mount the MCP server, and ship workflows where every answer carries its source, verification, and audit trail.

Get an API key Browse the API surface

Attorney review required before filing, sending, or signing.