Open-Source MCP Gateway: Real-Time Auto-Blocking for Compromised AI Agents
LawFamiliar3588 · reddit · 2026-08-10
A developer open-sourced wardline, an MCP (Model Context Protocol)-aware proxy gateway designed to mitigate runtime security risks for AI agents.
Key Features:
- Comprehensive Interception: All tools/call, resources/, and prompts/ requests pass through policy, budget, and audit checks.
- Real-Time Auto-Blocking: Features a per-identity anomaly detector that can auto-block a client exhibiting behavioral spikes (e.g., if hijacked or maliciously prompted) without human intervention.
- Slow-and-Low Defense: Catches abrupt deviations while a baseline mechanism adapts to slow ramps, defending against low-and-slow exfiltration attacks.
Tech Stack: Supports three policy backends (YAML, OPA, Cedar) and is packaged as a single Go binary. The author is seeking community feedback on its threat model design.
More from coding & agent
- DIRECT: retrieval-based next-question suggestions for conversational agents — burkov · 2026-10-16
- Claude Code and Codex now autocomplete the user's next action — tokenbender · 2026-10-16
- Hermes Agent ships official Reverse Engineer Anything plugin with one-command setup — Teknium · 2026-10-16
- MLExpert: 35B MoE Agent Trained to Autonomously Handle Machine Learning Engineering — Shangheng Du · 2026-10-16
- Cowork-Gym: Alibaba Synthesizes 64.9K Agent Work Environments, Boosting Qwen by 8.7 Points — Anni Zou · 2026-10-16
- MMSearch-OPD: Joint RL and On-Policy Distillation for Long-Horizon Multimodal Search Agents — PekingUniversity · 2026-10-16