TypeScript SDK for autonomous Solidity security auditing — 7 breach scenarios, LangChain.js + Vercel AI SDK integration

I've been working on an autonomous security auditing gateway for Solidity smart contracts. The TypeScript SDK just went live on npm — it runs 7 automated breach simulations against any Solidity source code in ~2ms, with no signup or API key required for the free tier. What problem does this solve? Most Solidity security tools are either: Paid SaaS (Slither Pro, MythX) — requires account, subscription, API key CLI-only (Slither, Mythril) — requires local installation, Python/Foundry setup Manual — human auditors, weeks of lead time This SDK gives you a programmatic, agent-native audit that runs 7 breach scenarios via a single function call. No local toolchain. No signup. No API key for the free tier. Install npm install nexus-gateway-sdk The 7 Breach Scenarios ID Scenario Question BS-001 Unauthorized Minting Can tokens be minted without authorization? BS-002 Transfer Violation Can transfers bypass balance/allowance checks? BS-003 Fund Drain Can funds be withdrawn without authorization? BS-004 Emergency Freeze Is there an emergency stop / pause mechanism? BS-005 Ownership Renounce Can ownership renunciation lock admin functions? BS-006 Reentrancy Attack Are external calls protected against reentrancy? BS-007 Replay Attack Is nonce-based replay protection implemented? Each scenario returns a structured result: risk_level, affected_functions, detected, mitigation. Usage — 3 lines import { NexusClient } from "nexus-gateway-sdk"; const client = new NexusClient({ clientId: "my-app" }); const result = await client.dryRun({ source_code: contractSource }); // result.breach_simulation.overall_risk → "high" | "medium" | "low" | "critical" // result.deployable → true | false // result.breach_simulation.scenarios → array of 7 scenario objects // result.digital_twin_v3_matrix → clause-to-code mapping // result.urgency_signal → time-decay warning for M2M orchestrators Agent Framework Integration The SDK ships with native tool wrappers for AI agent frameworks: // LangChain.js import { createNexusTools } from "nexus-gateway-sdk"; const tools = createNexusTools({ clientId: "my-agent" }); // → Returns Tool[] compatible with LangChain.js agent executor // Vercel AI SDK import { createNexusVercelTools } from "nexus-gateway-sdk"; const tools = createNexusVercelTools({ clientId: "my-agent" }); // → Returns tools compatible with Vercel AI SDK generateText This means any LangChain.js or Vercel AI agent can autonomously audit Solidity contracts as part of its toolset — no custom integration needed. Digital Twin v3.1 Matrix Beyond the 7 breach scenarios, the audit produces a Digital Twin matrix — a clause-to-code mapping that links contract functions to their breach conditions: // Each function in the contract gets mapped: { function_signature: "withdraw()", visibility: "external", modifiers: [], breach_conditions: ["BS-006: reentrancy via external call before state update"] } This is designed for M2M orchestration — agents can programmatically assess which functions are safe to call and which require additional guards. Production Verification The gateway passed 17/17 self-validation checks: ✅ All 7 gateway endpoints live ✅ ERC-8004 on-chain identity (Agent #636, Polygon Mainnet) ✅ EIP-712 USDC pull payment contract deployed ✅ npm + PyPI packages live ✅ OpenAPI 3.0.3 spec with M2M x-extensions ✅ A2A discovery manifest (.well-known/agent.json) ✅ Treasury wallet funded ✅ USDC contract verified Try It Now — No Install Interactive playground: https://rakhmadaa-gif.github.io/nexus-core-gateway/ Paste any Solidity code → click "Run Security Audit" → instant results with 7 breach scenarios, risk badges, and recommendations. Or via curl: curl -X POST https://xibzsthfrbomefnvbicb.supabase.co/functions/v1/hello-world/gateway/dry-run \ -H 'Content-Type: application/json' \ -d '{"source_code": "pragma solidity ^0.8.20; contract Token { }"}' SDK Stats npm: nexus-gateway-sdk@1.0.0 — 136 downloads PyPI: nexus-gateway-sdk@1.0.0 — 498 downloads License: MIT Latency: ~2ms per audit SLA: