Add Lexius Compliance MCP server

servers/lexius/readme.md

@@ -0,0 +1,59 @@
+# Lexius Compliance MCP Server
+
+AI regulatory compliance database with provenance-tracked obligations, penalties, deadlines, and verbatim regulation text.
+
+## What it does
+
+Lexius gives Claude access to a structured compliance database covering the **EU AI Act** (Regulation 2024/1689) and **DORA** (Regulation 2022/2554). Every fact is tagged with a provenance tier:
+
+- **AUTHORITATIVE** — verbatim from the official regulation text (EUR-Lex CELLAR), SHA-256 hash-verified
+- **CURATED** — written or reviewed by a domain expert, with attribution
+- **AI_GENERATED** — model output, flagged for review
+
+## Tools (13)
+
+| Tool | What it does |
+|------|-------------|
+| `legalai_classify_system` | Classify an AI system's risk level |
+| `legalai_get_obligations` | Get compliance obligations by role/risk |
+| `legalai_calculate_penalty` | Calculate fine exposure (EUR 35M/15M/7.5M tiers) |
+| `legalai_get_article` | Retrieve verbatim article text |
+| `legalai_get_deadlines` | Compliance deadlines with days remaining |
+| `legalai_search_knowledge` | Semantic search across all content |
+| `legalai_answer_question` | FAQ lookup |
+| `legalai_run_assessment` | Structured assessments (Art. 6(3), GPAI) |
+| `legalai_list_legislations` | Available legislations |
+| `legalai_get_article_history` | Article revision history |
+| `legalai_get_derivation_chain` | Trace obligation → source articles |
+| `legalai_get_article_extracts` | Extracted facts (fines, dates, cross-refs) |
+| `legalai_run_swarm_assessment` | Parallel hivemind compliance assessment |
+
+## Modes
+
+### Direct Mode (own database)
+
+Connect to a PostgreSQL database with the Lexius schema. Use the `robotixai/lexius-db` Docker image for a pre-configured database:
+
+```bash
+docker run -d -p 5432:5432 -e POSTGRES_PASSWORD=secret robotixai/lexius-db
+```
+
+Then configure the MCP server with `DATABASE_URL=postgresql://legal_ai:secret@host.docker.internal:5432/legal_ai`.
+
+### Proxy Mode (hosted API)
+
+Connect to a hosted Lexius API — no database required. Configure with `LEXIUS_API_URL` and `LEXIUS_API_KEY`.
+
+## Data
+
+- **190 articles** (113 EU AI Act + 13 annexes + 64 DORA) — all AUTHORITATIVE
+- **1,734 extracted facts** (fine amounts, percentages, dates, cross-references, shall-clauses)
+- **61 curated obligations** (35 EU AI Act + 26 DORA)
+- **5 penalty tiers** with CI-verified amounts
+- **44 FAQ entries** with semantic search
+
+## Links
+
+- [GitHub](https://github.com/rob-otix-ai/lexius)
+- [npm: @robotixai/lexius-mcp](https://www.npmjs.com/package/@robotixai/lexius-mcp)
+- [Docker Hub: robotixai/lexius-mcp](https://hub.docker.com/r/robotixai/lexius-mcp)

servers/lexius/server.yaml

@@ -0,0 +1,58 @@
+name: lexius
+image: mcp/lexius
+type: server
+meta:
+  category: database
+  tags:
+    - compliance
+    - regulation
+    - legal
+    - eu-ai-act
+    - dora
+    - provenance
+about:
+  title: Lexius Compliance
+  description: >-
+    AI regulatory compliance database with provenance-tracked obligations,
+    penalties, deadlines, and verbatim regulation text. Covers the EU AI Act
+    and DORA. Every fact is labelled AUTHORITATIVE (verbatim from EUR-Lex),
+    CURATED (expert-reviewed), or AI_GENERATED. Includes deterministic fact
+    extraction and parallel hivemind swarm assessment.
+  icon: https://avatars.githubusercontent.com/u/209498797?v=4
+source:
+  project: https://github.com/rob-otix-ai/lexius
+  branch: main
+  commit: d50a452d8c6ac1214fd2ad281d396ab9c757b893
+  directory: packages/mcp
+config:
+  description: >-
+    Direct mode: set DATABASE_URL to a PostgreSQL database with the Lexius
+    schema (use 'docker run robotixai/lexius-db' for a ready-made database).
+    Proxy mode: set LEXIUS_API_URL and LEXIUS_API_KEY to connect to a hosted
+    Lexius API without needing a local database.
+  secrets:
+    - name: lexius.api_key
+      env: LEXIUS_API_KEY
+      example: lx_your_api_key_here
+  env:
+    - name: DATABASE_URL
+      example: postgresql://legal_ai:secret@host.docker.internal:5432/legal_ai
+      value: "{{lexius.database_url}}"
+    - name: LEXIUS_API_URL
+      example: https://api.lexius.ai
+      value: "{{lexius.api_url}}"
+    - name: LEXIUS_API_KEY
+      example: lx_your_api_key_here
+      value: "{{lexius.api_key}}"
+  parameters:
+    type: object
+    properties:
+      database_url:
+        type: string
+        description: PostgreSQL connection string (direct mode)
+      api_url:
+        type: string
+        description: Hosted Lexius API URL (proxy mode)
+      api_key:
+        type: string
+        description: API key for hosted Lexius API (proxy mode)

servers/lexius/tools.json

@@ -0,0 +1,108 @@
+[
+  {
+    "name": "legalai_classify_system",
+    "description": "Classify an AI system under a legislation's risk framework (e.g., EU AI Act prohibited/high-risk/limited/minimal). Uses signals, keywords, and semantic matching.",
+    "arguments": [
+      { "name": "legislationId", "type": "string", "desc": "Legislation ID (e.g., 'eu-ai-act', 'dora')" },
+      { "name": "description", "type": "string", "desc": "Free-text description of the AI system" },
+      { "name": "role", "type": "string", "desc": "Organization's role: 'provider', 'deployer', or 'unknown'" }
+    ]
+  },
+  {
+    "name": "legalai_get_obligations",
+    "description": "Get compliance obligations filtered by legislation, role, and risk level. Each obligation includes its provenance tier (AUTHORITATIVE/CURATED).",
+    "arguments": [
+      { "name": "legislationId", "type": "string", "desc": "Legislation ID" },
+      { "name": "role", "type": "string", "desc": "Filter by role (e.g., 'provider', 'deployer')" },
+      { "name": "riskLevel", "type": "string", "desc": "Filter by risk level (e.g., 'high-risk', 'limited')" }
+    ]
+  },
+  {
+    "name": "legalai_calculate_penalty",
+    "description": "Calculate potential penalties for a specific violation type. Returns AUTHORITATIVE fine amounts extracted from verbatim regulation text.",
+    "arguments": [
+      { "name": "legislationId", "type": "string", "desc": "Legislation ID" },
+      { "name": "violationType", "type": "string", "desc": "Violation type (e.g., 'prohibited-practices', 'high-risk-non-compliance')" },
+      { "name": "annualTurnoverEur", "type": "number", "desc": "Annual worldwide turnover in EUR" },
+      { "name": "isSme", "type": "boolean", "desc": "Whether the organization is an SME" }
+    ]
+  },
+  {
+    "name": "legalai_get_article",
+    "description": "Retrieve a specific article by number from a legislation. Returns verbatim regulation text (AUTHORITATIVE) when fetched from EUR-Lex CELLAR.",
+    "arguments": [
+      { "name": "legislationId", "type": "string", "desc": "Legislation ID" },
+      { "name": "articleNumber", "type": "string", "desc": "Article number (e.g., '6', '99', 'annex-iv')" }
+    ]
+  },
+  {
+    "name": "legalai_get_deadlines",
+    "description": "Get compliance deadlines for a legislation with days remaining and past/upcoming status.",
+    "arguments": [
+      { "name": "legislationId", "type": "string", "desc": "Legislation ID" },
+      { "name": "onlyUpcoming", "type": "boolean", "desc": "If true, only return future deadlines" }
+    ]
+  },
+  {
+    "name": "legalai_search_knowledge",
+    "description": "Semantic search across articles, obligations, FAQs, or risk categories. Results include provenance tier.",
+    "arguments": [
+      { "name": "legislationId", "type": "string", "desc": "Legislation ID" },
+      { "name": "query", "type": "string", "desc": "Natural language search query" },
+      { "name": "entityType", "type": "string", "desc": "Type: 'article', 'obligation', 'faq', 'risk-category'" },
+      { "name": "limit", "type": "number", "desc": "Max results (default 5)" }
+    ]
+  },
+  {
+    "name": "legalai_answer_question",
+    "description": "Answer a compliance question using the FAQ knowledge base with semantic matching.",
+    "arguments": [
+      { "name": "legislationId", "type": "string", "desc": "Legislation ID" },
+      { "name": "question", "type": "string", "desc": "The question to answer" }
+    ]
+  },
+  {
+    "name": "legalai_run_assessment",
+    "description": "Run a structured compliance assessment (e.g., Article 6(3) exception check, GPAI systemic risk).",
+    "arguments": [
+      { "name": "legislationId", "type": "string", "desc": "Legislation ID" },
+      { "name": "assessmentId", "type": "string", "desc": "Assessment identifier" }
+    ]
+  },
+  {
+    "name": "legalai_list_legislations",
+    "description": "List all available legislations in the database (currently EU AI Act and DORA).",
+    "arguments": []
+  },
+  {
+    "name": "legalai_get_article_history",
+    "description": "Retrieve the full revision history of an article, showing how verbatim text changed across amendments.",
+    "arguments": [
+      { "name": "articleId", "type": "string", "desc": "Article identifier (e.g., 'eu-ai-act-art-99')" }
+    ]
+  },
+  {
+    "name": "legalai_get_derivation_chain",
+    "description": "Trace an obligation back to its source articles. Shows the verbatim law that a curated obligation paraphrases.",
+    "arguments": [
+      { "name": "obligationId", "type": "string", "desc": "Obligation identifier" }
+    ]
+  },
+  {
+    "name": "legalai_get_article_extracts",
+    "description": "View deterministically extracted facts from an article's verbatim text: fine amounts, percentages, dates, cross-references, shall-clauses. All AUTHORITATIVE.",
+    "arguments": [
+      { "name": "articleId", "type": "string", "desc": "Article identifier (e.g., 'eu-ai-act-art-99')" },
+      { "name": "extractType", "type": "string", "desc": "Optional filter: 'fine_amount_eur', 'turnover_percentage', 'date', 'article_cross_ref', 'shall_clause'" }
+    ]
+  },
+  {
+    "name": "legalai_run_swarm_assessment",
+    "description": "Run a parallel hivemind assessment — multiple agents analyse all articles simultaneously, producing a comprehensive ComplianceReport with gap detection. Fully deterministic (no LLM in the agent loop).",
+    "arguments": [
+      { "name": "legislationId", "type": "string", "desc": "Legislation ID (e.g., 'eu-ai-act')" },
+      { "name": "systemDescription", "type": "string", "desc": "Description of the AI system being assessed" },
+      { "name": "concurrency", "type": "number", "desc": "Number of parallel agents (default 4, max 8)" }
+    ]
+  }
+]