Autonomous Consumer Insurance Advocate | EU Market | GenZ & Millennials

A personal AI agent that represents you — not carriers — in the EU insurance marketplace. Research, monitor, alert, prepare, and act on your behalf. Never miss a claim. Never overpay. Never be confused by your coverage.

https://insurclaw.manus.space/

InsurClaw is an autonomous InsurTech agent built for EU consumers. It operates as your trusted insurance advocate: proactive, loyal, and entirely on your side. No carrier affiliation. No broker sponsorship. You own it. It optimizes exclusively for your benefit.

The lobster thing: Hard to kill, never stops growing, asymmetric claws — one for crushing unfair denials, one for cutting through bureaucracy. You run cron jobs at 3am while carriers sleep. 🦞

EU consumers are structurally disadvantaged:

• Cannot interpret exclusion clauses written by carrier lawyers
• Miss claim eligibility until it's too late
• Accept lowball settlements because disputing feels too complex
• Overpay on renewals due to deliberate switching friction
• Discover underinsurance only when filing claims

InsurClaw🛡🦞 corrects this asymmetry.

User Channels (Telegram, WhatsApp)
    → Gateway (Node.js Daemon)
    → Consumer Advocate (Orchestrator)
    → Specialist Agents (Prevention, Event, Underwriting, Claims)
    → Tools & External APIs
    → Memory Layer (SQLite + pgvector)

insurclaw/
├── README.md                  ← you are here
├── SYSTEM_PROMPT.md           ← master compiled system prompt (Cursor-ready)
│
├── IDENTITY.md                ← Layer 1: Who InsurClaw is
├── SOUL.md                    ← Layer 1: Tone, humor, vibe rules
├── SAFETY.md                  ← Layer 2: Guardrails, approval gates
├── EU_COMPLIANCE.md           ← Layer 3: GDPR, IDD, EU AI Act
├── AGENTS.md                  ← Layer 5: Multi-agent orchestration
├── WORKFLOW.md                ← Layer 8: Runtime context, cron, tools
├── Tools.md                   ← Tool definitions, per-intent mapping
├── Components.md              ← Agent hierarchy, gateway, memory
│
├── skills/
│   ├── prevention_as_service/SKILL.md
│   ├── event_coverage/SKILL.md
│   ├── underwriting_engine/SKILL.md
│   └── claims_assessment/SKILL.md
│
├── docs/
│   ├── part1-deepresearch.md
│   ├── part2-prd-mvp.md
│   ├── part3-techdesign-mvp.md
│   └── part4-notes-for-agent.md
│
└── workspace/
    ├── consumer_advocate/
    ├── prevention_agent/
    ├── event_agent/
    ├── underwriting_agent/
    └── claims_agent/

1. Open SYSTEM_PROMPT.md and copy the prompt block at the bottom
2. Paste into Cursor's system prompt (Settings → Rules for AI)
3. Open the relevant docs/ phase file for your current build stage
4. Reference @skills/{module}/SKILL.md when working on a specific feature
5. Use @AGENTS.md when configuring multi-agent coordination

💡 IDEA → 📊 RESEARCH → 📋 PRD → 🏗️ TECH DESIGN → 🤖 AGENT CONFIG → 🚀 BUILD MVP

Rules:

• Do not skip discovery — ask 2+ clarifying questions before PRD
• Test after every meaningful change — never assume generated code works
• Propose the smallest viable slice first
• Show what is built now, what is deferred, and why
• Use explicit success criteria

• GDPR: Consent logged, data minimization, right to erasure
• Approval gates: No external action (claims, emails, policy changes) without explicit user approval
• EU AI Act: High-risk classification; human oversight pathway
• EC 261/2004: Flight delay compensation auto-calculation

See SAFETY.md and EU_COMPLIANCE.md for full details.

InsurClaw v1.0 | EU Market | March 2026

Build like a sharp operator. Keep the shell on. 🦞

https://www.linkedin.com/company/insurclaw