# Claude Startups — Ready-to-Paste Answers

## Version A — Short (ultra ringkas)
**What are you building?**
AuditAI is an AI-powered smart contract security copilot for Web3 teams. It combines static analysis and LLM reasoning to detect vulnerabilities and return exploit-aware, developer-ready fixes in a structured format.

**Who is it for?**
Early-stage Web3 teams shipping Solidity contracts and needing fast pre-audit security feedback.

**Why now?**
Smart contracts ship faster than security review cycles. Teams need lightweight, continuous security checks before paying for full audits.

**Current traction/progress?**
Working MVP is live with end-to-end scanning. Internal benchmark on 5 representative contracts correctly identified critical/medium issues and returned 0 findings for a safe baseline contract.

**How will Claude support help?**
Model credits and reliability feedback will accelerate quality iteration, false-positive reduction, and integration into developer workflows.

---

## Version B — Medium (balanced)
**What are you building?**
We’re building AuditAI, an AI smart contract security copilot. Developers paste Solidity contracts and get a structured report with security summary, severity-ranked findings, exploit scenario, and concrete remediation guidance.

**What problem are you solving?**
Most early Web3 teams can’t run deep security review on every iteration. Existing scanners are noisy and hard to prioritize. This creates a gap between coding speed and security confidence.

**Who is the target user?**
- Early-stage DeFi/protocol teams
- Solo smart contract founders
- Engineering teams doing pre-audit triage

**How is your approach different?**
AuditAI focuses on exploitability and practical fixes, not just detector output. We also keep output schema stable for downstream automation (reporting, CI checks, and internal security workflows).

**What have you validated so far?**
MVP pipeline is running in production-like setup:
- frontend UI + backend API
- Slither integration
- Anthropic-compatible LLM integration via router
- fallback mode (Slither-only) when model output fails

Internal benchmark (5 contracts):
- Reentrancy: detected (Critical)
- Access control flaw: detected (Critical)
- Unchecked call: detected (Medium)
- Oracle misuse: multiple findings (Critical/High/Medium)
- Safe baseline: no findings

**What do you plan in the next 8 weeks?**
- Finding precision improvements
- Exportable reports (JSON/PDF)
- Project workspace + scan history
- CI/PR integration
- Pilot with 3–5 Web3 teams

**How can Claude Startups help?**
We need model credits and product feedback to improve reasoning quality, reliability, and latency while we onboard pilot teams.

---

## Version C — Long (deeper)
**Company/Product Overview**
AuditAI is an AI-powered smart contract security copilot that helps Web3 teams catch critical vulnerabilities earlier in development. It combines deterministic static analysis with LLM reasoning to provide results developers can act on immediately.

**Core user workflow**
1. Developer pastes Solidity code in AuditAI.
2. Backend runs static analysis.
3. AI layer synthesizes findings into structured output.
4. User receives:
   - summary of security posture
   - severity-ranked findings
   - exploit scenario for each significant issue
   - concrete fix guidance

**Problem and urgency**
Smart contract teams move fast, but security review is expensive and bottlenecked. In practice, many teams either ship with limited review or delay releases. Existing tools often return noisy detector output that still requires expert interpretation.

**Why AuditAI now**
The current wave of L2/Web3 product iteration demands a fast “first-pass” security layer that is always available during development, not only at final audit stage.

**Differentiation**
- exploit-aware explanations instead of raw alerts
- remediation-focused output for faster dev action
- stable structured schema for integrations
- fallback reliability mode when model output is unavailable

**Current state / validation**
We have a working MVP with UI/API and reproducible benchmark artifacts. Internal tests on five representative contracts validated both high-risk detection and low-noise behavior on a safe baseline.

**Roadmap and go-to-market**
Near-term:
- export/reporting
- workspace/team features
- CI/PR integrations
- pilot programs with Web3 teams

Go-to-market starts with founder-led onboarding for security-conscious protocol teams, then expands through CI integration and shared report workflows.

**What we need from Claude Startups**
- model credits for rapid product iteration
- feedback on prompt/reliability architecture
- support while moving from MVP to pilot readiness