Security bodyguard for AI trading agents.
SolShield MCP does not trade on your behalf — it protects autonomous trading agents before they sign. Connect Claude, OpenClaw, Cursor, or custom execution bots to catch honeypots, upgradeable proxies, phishing domains, and prompt-injected token metadata.
Live Agent Guardrail Simulator
Run a pre-trade simulation to inspect the agent verdict
Returns deterministic ALLOW, WARN, or BLOCK output along with sanitized agent-safe metadata and on-chain RPC proofs.
Connect via Streamable HTTP MCP
{
"mcpServers": {
"solshield": {
"type": "http",
"url": "https://sol-shield.xyz/api/mcp"
}
}
}Exposed MCP Tools (6 Read-Only Guardrails)
JSON-RPC 2.0Combines prompt-injection filtering, domain checks, and live RPC bytecode/authority simulation into ALLOW, WARN, or BLOCK.
Strips role overrides, invisible Unicode BiDi stealth traps, base64-smuggled prompts, and wallet-drain imperatives.
Checks EIP-1967 proxies, Sourcify source, GoPlus/Honeypot.is sell tax across 7 EVM chains plus Solana mint/freeze authority.
Scans an agent's public address for lingering unlimited ERC-20 / Permit2 spenders or Solana SPL delegates.
Checks live balances and recent on-chain dumps of suspected MM, early buyer, or deployer wallets.
Blocks typosquatted DEX clones and phishing frontends before an autonomous browser agent signs typed data.