luckyPipewrench/pipelock
★ 938 stars
Covered in week 2026-W19 · the 2026-05 report
Open-source AI agent firewall for MCP security: agent egress control, DLP, SSRF, and prompt injection defense.
What pipelock is
Pipelock sits between AI agents and the network. It inspects mediated HTTP, WebSocket, MCP, and A2A traffic, plus CONNECT tunnel contents when TLS interception is enabled, for secret exfiltration, prompt injection, SSRF, tool poisoning, and risky tool-call chains. Plain CONNECT without interception is scanned at the hostname and URL level. Configured MCP upstreams are an exception to private-address SSRF blocking: local/private servers are allowed, but cloud metadata endpoints remain blocked.
From the repository README · read it on GitHub →
Transcript
Enterprise AI firewalls like Lakera and Prompt Security cost over $1,000 monthly. PipeLock does it free. This open-source AI agent firewall stops data leaks, SSRF attacks, and prompt injection attempts at the network layer. 549 GitHub stars are already using it to secure their MCP deployments. With zero-cost agent egress control and enterprise-grade DLP built in, you can protect your AI stack without paying five figures annually for proprietary solutions. Star it if this saved you money.