I built Agent Shield, an open-source traffic control layer for AI coding agents.

Agent Shield is an open-source proxy layer that sits between AI coding agents and the network to inspect, control, and filter outbound traffic.

It can monitor, audit, block, redact, reroute, and route HTTP, WebSocket, and SSE traffic, with alerts and dashboards for visibility.

The system adds centralized policy enforcement for agent communications, helping reduce leaks, control costs, and coordinate multiple agents.

It supports integrations such as MITM proxying, Ollama, llama.cpp, NATS, and DLP services, without relying on vendor-specific clients.