SolGuard: On-chain Token Safety for Solana AI Trading Agents

If you're building autonomous trading agents on Solana, you've hit this problem: your agent needs to vet a token before it buys — but most agents trade blind. They have no idea whether a token is a rug pull, a honeypot, or has an active mint authority that can dilute holders to zero. I built SolGuard to fix this: an MCP server that screens any SPL / Token-2022 token for the common scam patterns, then executes the swap through an MEV-protected route — screening and execution in one place. Mint & freeze authority — is supply still mintable? Can your tokens be frozen in place? Honeypot detection — can the token actually be sold, or only bought? Liquidity & holders — concentration, LP locks, real depth. Token-2022 traps — permanent delegate, transfer hooks, default-frozen state. verify_token_safety — full on-chain safety audit check_authorities — mint / freeze / Token-2022 trap detection simulate_sell — honeypot check execute_safe_swap — MEV-protected execution SolGuard is listed on the official MCP Registry as io.github.MrWizardlyLoaf/solguard. Remote endpoint, no install: https://web-production-58d585.up.railway.app/mcp Or self-host — it's open source (MIT) on GitHub: https://github.com/MrWizardlyLoaf/solguard-mcp A rug pull on Solana often looks fine at the moment of purchase — liquidity is there, the chart is green. The trap is in the authority: a live mint authority lets the deployer print unlimited supply after you buy; a freeze authority can lock your tokens so you can never sell. An autonomous agent that doesn't read these on-chain fields walks straight into it. SolGuard reads them directly and gives a verdict on a fresh launch instead of unknown. Built for Solana trading agents. Feedback welcome.