AILegal — verified legal provisions for agents

AILegal MCP server

Web product: https://ailegal.legal — same corpus, same refusal when no provision is verified.

MCP tool for agents that need an answer grounded in law, not in a model's memory.

legal_breakdown sends a situation to the public AILegal API and returns the provisions that apply, each with a link to the official text. When the verified corpus has no provision for the question, the tool returns an honest refusal (no_norms) instead of inventing an article number.

Jurisdictions: Russia (ru), India (in), the US (us), the UK (uk), Canada (ca), Hong Kong (hk). 20 India situation pages are available in Hindi in the web product.

Install

pip install "mcp>=2"

Claude Desktop / Cursor / Continue

{
  "mcpServers": {
    "ailegal": {
      "command": "python",
      "args": ["/absolute/path/ailegal_server.py"]
    }
  }
}

Tool

{"text": "My employer has not paid my salary in India", "jurisdiction": "in", "locale": "en"}

Response: {answer, norms: [{code, article, title, source_url, verified}], jurisdiction, refused, coverage}.

Why it refuses

A citation that does not exist is worse than no answer: the reader stops looking for the right provision. The server is deliberately limited to the verified corpus.

Links

• Product: https://ailegal.legal
• Example report: https://ailegal.legal/sample
• Legal information, not legal advice.

Related evidence products (same team)

• Deny evidence — a signed, append-only log of refusals (who asked, which policy, why): https://attest.rta.lat/deny-evidence
• EU AI Act evidence pack — audit log, deny log, human oversight, model card, risk assessment and data-flow map, from $1,500: https://attest.rta.lat/eu-ai-act-evidence
• Agent Spend Audit — five days on one month of agent logs, $500 in USDC: https://rta.lat/audit/ (note the domain: rta.lat/audit/, not attest.rta.lat/audit)