Pipelock: Open-source AI Agent Firewall
Help Net Security, Monday, May 4th, 2026
Pipelock is an open-source security proxy that enforces network isolation for AI coding agents to prevent credential leakage.
Pipelock is an open-source security harness developed by Joshua Waldrep that addresses vulnerabilities in AI coding agents by inserting an enforcement layer between agents and the network. Unlike traditional agent-security tools that rely on agent cooperation, Pipelock sits outside the agent at the egress boundary, similar to how TLS handles trust on the web.
The system runs as a lightweight Go binary with 22 dependencies under Apache 2.0 license and features an 11-layer scanning pipeline covering credential protection, SSRF prevention, data loss prevention, and injection detection with 48 credential patterns and 25 injection detection patterns.
It supports multiple protocols including HTTP, WebSocket, Model Context Protocol, and provides tamper-evident audit logging with Ed25519 signatures, compliance mappings to OWASP and NIST standards, and SARIF integration for GitHub Code Scanning.