LeakGuard lite: find and redact leaked secrets before an agent sends text

Overview

LeakGuard: outbound leak check for AI agents. Before an agent sends text to an LLM, tool, log or webhook, it detects and redacts common secret formats: BIP39 seed phrases (checksum-verified), EVM and Solana private keys, WIF, xprv, API keys (GitHub, AWS, Stripe, OpenAI, Anthropic, Slack and more), JWT, PEM and PII. Lite ($0.005, POST {"text": ...}, up to 32 KB): verdict (WALLET_KEY_EXPOSED / SECRETS_FOUND / PII_FOUND / CLEAN), safe_to_send, findings with type, severity, offsets and a masked preview (never the secret), a remediation line per finding, and the redacted text. Limits: pattern + checksum detection (no ML), so it won't catch every secret and novel formats can be missed. Inputs are processed in memory for the request only, not logged or stored. USDC on Base, Polygon, Arbitrum One or Solana via x402, or prepaid YourAITools credit keys; no signup. Built by YourAITools (youraitools.co).

Health