Structured case law, statutes, and regulations — delivered natively through the Model Context Protocol. One connection, every model. Every answer citable to the paragraph.
POST https://mcp.lawmcp.ai/ { "method": "tools/call", "params": { "name": "find_rulings", "arguments": { "jurisdiction": "IT", "topic": "GDPR · art.32", "since": "2023-01-01", "cite": true } }, "jsonrpc": "2.0", "id": 1 }
The models are good. The interfaces are good. The data isn't.
Court archives are inconsistent. Statutes drift across versions. Citations are unresolved strings. PDFs don't compose into reasoning. So every team building a legal AI product ends up rebuilding the same data stack from scratch — burning a year before they ship a single feature.
LawMCP fixes the data problem once. Structured corpus, paragraph-level citations, agent skills — delivered natively through MCP. Build on top instead of inside.
Structured legal data is necessary but not enough. Agents need pre-built reasoning patterns — the things lawyers actually do with the data. LawMCP ships both, designed to compose.
Norms, decisions, regulations, and doctrine — parsed, normalized, and addressable down to the paragraph. Citations are resolved edges, not strings.
Pre-built skills that turn the corpus into action: find, compare, summarize, verify. Compose them into agent workflows in minutes, not months.
find_rulings → find_norms → calculate_deadline in one agent loop.Honest answer: Italy and the European Union are both live and complete today — 7M+ legal documents under one ontology. Germany is rolling. Spain, France, Austria, and Switzerland are in build. Every jurisdiction lands with the same skills, the same citation guarantees, and no rewrites when you cross a border.
Each skill is an MCP tool with typed inputs, typed outputs, and citations by default. Four families — retrieve the law, analyze it against a fact pattern, compose the work product, and compute procedure and timing. The essentials are below, all live — more ship every week. Chain them into agent loops or call them directly.
Search case law across jurisdictions — filter by court, topic, and date. Returns ranked rulings with paragraph-level snippets and stable citations.
Search statutes, codes, and regulations — national and regional. Returns normative text with version metadata, in-force status, stable IDs, and links to amending acts.
Search legal literature — journals, treatises, and commentary. Returns ranked excerpts with provenance, linked to the norms and rulings they discuss.
Resolve any citation into a structured reference and traverse the citation graph in either direction — what it cites, what cites it — to any depth.
Check a fact pattern against current law — applicable norms, supporting and conflicting authorities, and a structured risk read.
Analyze a contract — extract clauses, obligations, and risks, and flag missing protections against applicable law.
Review any legal document — classify it, extract key terms and dates, and surface issues and inconsistencies.
Generate a litigation or transaction strategy for a matter — issues, options, and a recommended path, each backed by authority.
Draft a pleading from your facts and the retrieved authorities — structured, formatted, and citation-backed on every assertion.
Generate a contract or clause from your terms and a template, aligned to current law in the target jurisdiction.
Compare a document against a reference standard or prior version and return a structured redline with rationale for each change.
Structured summary of a decision — facts, holding, ratio, dissent — each labeled and cited to its source paragraph.
Compute a procedural deadline from a triggering event — service, hearing, notification — accounting for suspensions, holidays, and procedural rules.
Compute the full deadline cascade for a pleading sequence from a hearing date — every intermediate term, ordered.
LawMCP speaks the Model Context Protocol from day one — and only the Model Context Protocol. Connect once, get every skill with typed schemas, full streaming, and built-in error handling. No SDKs to wrap, no glue code to maintain, no protocol-mismatch tax.
No bespoke retrievers. No glue code. No "wrap our SDK in a tool definition" tickets. Drop the MCP endpoint into your agent runtime and every skill is callable in the next prompt.
{
"mcpServers": {
"lawmcp": {
"type": "http",
"url": "https://mcp.lawmcp.ai/",
"headers": {
"Authorization": "Bearer sk_live_..."
}
}
}
}
curl -X POST "https://mcp.lawmcp.ai/" \ -H "Content-Type: application/json" \ -H "Authorization: Bearer sk_live_..." \ -d '{ "jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": { "name": "find_rulings", "arguments": { "jurisdiction": "IT", "topic": "GDPR · art.32", "since": "2023-01-01", "cite": true } } }'
A legal answer without its source is a guess. Every skill returns its citations alongside its result — court, date, paragraph, version, link. Built so your agent's output stands up in front of a partner, a client, or a regulator.
Under the GDPR, the Court of Justice of the EU held in Case C‑340/21 that the burden of proving the adequacy of technical and organisational security measures under art. 32 GDPR rests on the data controller — unauthorised third-party access does not, on its own, make those measures inadequate.
Law firms, banks, government, healthcare — the buyers who need legal AI most also need infrastructure that survives procurement. LawMCP is built to clear it.
Data residency in the EU. Sub-processor list, DPA, and ROPA available on request. Designed against the Italian Bar Association's confidentiality requirements out of the box.
Your queries and your agents' calls are never used to train models — ours or anyone else's. Logs are retained only for operational reliability, with configurable retention down to zero.
Every response is fully attributable. Replayable, traceable, and exportable for compliance review. Built so your agent's reasoning is defensible — not just convincing.
Three tiers, one corpus. Try the MCP for free, pay only when you ship, and switch to enterprise the day you need dedicated infrastructure, SSO, or a custom jurisdiction.
Build, prototype, and demo. Real corpus, real skills, generous quota.
Ship to customers. Pay per skill call, hosted MCP endpoint, SLA on response time.
For law firms, banks, government, healthcare. Dedicated, SSO-wired, white-glove.
Join the early-access wait list. Onboarding partners in waves as MCP capacity scales.