Open-source MCP firewall blocks prompt injection in 2.7ms with deterministic regex and AST parsing
EstablishmentTough18 · reddit · 2026-09-28
An open-source Security Gate x402 MCP server adds sub-3ms deterministic protection against indirect prompt injection and tool poisoning for autonomous agents that run bash, scrape the web, or sign on-chain transactions via Safe{Wallet}.
Instead of slow LLM-as-a-judge checks (500-2000ms per call), it uses 100% deterministic regex and AST parsing: 1.84ms for secret-leak/AST scans, 2.73ms for prompt injection deflection, and exactly 0 Gwei gas lost on intercepted attacks, which are stopped off-chain before EIP-712 signing. Ships with a ready-made Claude Desktop config, a one-line Python GuardedSafeWallet integration, and a live JSON-RPC demo endpoint; repo, PyPI package, and online inspector are public.
More from coding & agent
- Study of 2,170 GitHub projects maps how the fast, low-cost Jev decision model is used in the wild — CUHK-CSE · 2026-09-28
- Anthropic releases free 37-minute guide on building AI agents that automate entire businesses — Aiden_Tech_Ai · 2026-09-28
- Spotify paper: synthetic multi-turn dialogs and self-improvement loops boost conversational recsys quality by 8% — _reachsumit · 2026-09-28
- RecToolBench: 1,200+ Task Benchmark Tests Recommender Agents on MCP Tool Orchestration — _reachsumit · 2026-09-28
- Grok Build v1.0.43 fixes MCP prompts going missing and hanging silently in minimal mode — XFreeze · 2026-09-28
- Dev Open-Sources Claude Code Skill That Makes Full AI Music Videos End-to-End — TimothyDuignan · 2026-09-28