pentagonal

# Pentagonal — Smart Contract Forge

> AI-powered smart contract creation, auditing, and deployment across 14 blockchain networks with 8 autonomous security agents.

## Overview

Pentagonal is a sovereign smart contract platform that generates, audits, and deploys Solidity and Rust smart contracts. It uses 8 autonomous AI security agents working in parallel to pen-test every contract before deployment. The platform supports both EVM chains (Ethereum, Polygon, BSC, Arbitrum, Base, Optimism, Avalanche) and non-EVM chains (Solana via Anchor/Rust).

Website: https://pentagonal.ai

---

## Architecture

### Multi-Agent Security System

Pentagonal employs 8 specialized AI agents that run simultaneously during every audit:

1. **Reentrancy Agent** — Detects external call vulnerabilities, cross-function reentrancy, and state manipulation patterns. Checks for check-effects-interactions violations and delegate call risks.

2. **Flash Loan Agent** — Identifies flash loan attack vectors including price manipulation, governance attacks, and liquidity pool exploitation. Analyzes oracle dependencies and atomic transaction risks.

3. **Access Control Agent** — Examines authorization logic, role-based access, ownership patterns, and privilege escalation paths. Validates modifier usage and admin function protection.

4. **Gas Optimization Agent** — Analyzes gas consumption patterns, identifies storage optimization opportunities, suggests calldata vs memory usage improvements, and flags unnecessary computations.

5. **Oracle Manipulation Agent** — Detects oracle dependency risks, TWAP manipulation vectors, stale price vulnerabilities, and multi-oracle inconsistency attack surfaces.

6. **MEV/Front-Running Agent** — Identifies transaction ordering dependencies, sandwich attack vectors, deadline manipulation risks, and mempool visibility exposures.

7. **Integer Overflow Agent** — Checks for arithmetic overflow/underflow vulnerabilities, unchecked math blocks, precision loss in division, and type casting issues.

8. **Economic Exploit Agent** — Analyzes tokenomics logic, fee calculation vulnerabilities, reward distribution flaws, and economic invariant violations.

### Self-Learning Security Engine

Every audit enriches Pentagonal's security knowledge base. The platform accumulates security rules and patterns from each scan, making future contract generation inherently safer. This creates a compounding security advantage over time.

---

## Supported Blockchain Networks

### EVM Mainnets (Solidity)
| Chain | Chain ID | Native Token |
|-------|----------|-------------|
| Ethereum | 1 | ETH |
| Polygon | 137 | MATIC |
| BNB Smart Chain | 56 | BNB |
| Arbitrum One | 42161 | ETH |
| Base | 8453 | ETH |
| Optimism | 10 | ETH |
| Avalanche C-Chain | 43114 | AVAX |

### Non-EVM Mainnets
| Chain | Runtime | Native Token |
|-------|---------|-------------|
| Solana | Anchor/Rust | SOL |

### Testnets
Sepolia, Amoy, Fuji, Arbitrum Sepolia, Base Sepolia, Optimism Sepolia, BSC Testnet, Solana Devnet

---

## Smart Contract Generation

### Solidity (EVM)
Pentagonal generates production-ready Solidity smart contracts from natural language descriptions. Contracts follow OpenZeppelin standards and include:
- SPDX license identifiers
- Pragma solidity version pinning
- NatSpec documentation
- Event emissions for all state changes
- Access control modifiers
- Reentrancy guards where applicable
- Gas-optimized storage patterns

### Rust/Anchor (Solana)
For Solana, Pentagonal generates Anchor framework programs including:
- Program derived addresses (PDAs)
- Account validation constraints
- Cross-program invocations (CPIs)
- SPL token integration
- Custom error types
- Event logging via `emit!`

---

## Integration Methods

### MCP Server (Model Context Protocol)
Pentagonal is available as an MCP server for AI IDE integration. Compatible with:
- Claude Desktop
- Cursor IDE
- Windsurf
- Any MCP-compatible client

**Available MCP Tools:**
- `pentagonal_lookup` — Token intelligence (free)
- `pentagonal_generate` — Generate smart contracts from natural language ($5)
- `pentagonal_audit` — Run 8-agent security scan on existing code ($5)
- `pentagonal_fix` — Auto-fix vulnerabilities found during audit (free)
- `pentagonal_compile` — Compile Solidity contracts server-side (free)
- `pentagonal_rules` — View accumulated security rules (free)
- `pentagonal_chains` — List all supported chains (free)

### Clawd Skill
Downloadable skill package for AI coding assistants with full contract generation, audit, fix, compile, and chain selection capabilities.

---

## Authentication Methods
- **Wallet Connect** — MetaMask, WalletConnect, Coinbase Wallet (EVM) + Phantom, Solflare (Solana)
- **Google OAuth** — Google account sign-in
- **GitHub OAuth** — GitHub account sign-in
- **X/Twitter OAuth** — X account sign-in
- **Email/Password** — Traditional email authentication

## Payment System
- Audits and contract generation cost $5 each
- Fixes and compilation are free for everyone
- Pay with EVM tokens (ETH, BNB, etc.), Solana (SOL), or stablecoins (USDC, USDT)
- AI agents can pay per-use with USDC on Base via x402 protocol (no account needed)
- Real-time price oracle via CoinGecko API

---

## API Reference

### Contract Generation
- **Endpoint:** POST /api/generate
- **Input:** Natural language description + target chain
- **Output:** Production-ready smart contract code with inline documentation
- **Credit cost:** $5 (1 credit)

### Contract Audit
- **Endpoint:** POST /api/audit
- **Input:** Smart contract source code + target chain
- **Output:** JSON audit report with severity ratings (Critical/High/Medium/Low/Informational), affected code ranges, and suggested fixes
- **Credit cost:** $5 (1 credit)

### Contract Fix
- **Endpoint:** POST /api/fix
- **Input:** Smart contract code + audit findings
- **Output:** Patched contract with all identified vulnerabilities resolved
- **Credit cost:** Free

### Contract Compile
- **Endpoint:** POST /api/compile
- **Input:** Solidity source code
- **Output:** ABI + bytecode (compiled via solc)
- **Credit cost:** Free

---

## FAQ

### What is Pentagonal?
Pentagonal is an AI-powered smart contract forge that creates, audits, and deploys smart contracts across 14 blockchain networks using 8 autonomous security agents.

### How does the multi-agent audit work?
When you submit a contract for audit, 8 specialized AI agents run in parallel, each analyzing the code from a different security perspective (reentrancy, flash loans, access control, gas optimization, oracle manipulation, MEV/front-running, integer overflow, and economic exploits). Results are aggregated into a comprehensive report.

### What chains does Pentagonal support?
EVM chains: Ethereum, Polygon, BSC, Arbitrum, Base, Optimism, Avalanche. Non-EVM: Solana (Anchor/Rust). Plus 8 testnets.

### How do credits work?
Security audits and contract generation each cost $5 per use (1 credit). Fixes and compilation are free for everyone. Credits are purchased with crypto. AI agents can pay per-use with USDC on Base via the x402 protocol — no account needed.

### Can I use Pentagonal in my AI IDE?
Yes. Pentagonal has a native MCP server that integrates with Claude Desktop, Cursor, Windsurf, and any MCP-compatible client. You can also download the Clawd Skill for direct AI assistant integration.

### Is my code stored?
Contract code is processed server-side for generation and auditing but is not permanently stored. Audit reports are associated with your account for reference.

### What makes Pentagonal different from other audit tools?
Three things: (1) Multi-agent architecture — 8 specialized agents vs. single-pass analysis. (2) Self-learning engine — security knowledge compounds across all audits. (3) Full lifecycle — generate, audit, fix, and deploy from one platform.

---

## Technical Stack
- **Frontend:** Next.js 15, React 19, TypeScript
- **Styling:** Vanilla CSS with design tokens
- **Auth:** Supabase Auth (OAuth + email + wallet signatures)
- **Payments:** On-chain verification (EVM via ethers.js, Solana via @solana/web3.js)
- **Wallet:** RainbowKit + wagmi (EVM), Solana Wallet Adapter
- **AI:** Multi-model architecture for specialized agent tasks
- **Deployment:** Vercel (Edge Functions)