OSS
Show HN: AgentGuard — Open-Source Firewall to Secure Autonomous AI Agents
Posted today on Show HN, AgentGuard is an open-source firewall that blocks dangerous tool calls before AI agents execute them — detecting SQL injection, path traversal, command injection, prompt injection, data exfiltration, and PII leakage. Self-hosted, MIT licensed, no telemetry, keeps no data outside user infrastructure. Deployable via Docker Compose with dashboard and gateway. Small but timely: as agent autonomy increases, the attack surface for tool-calling agents is a growing concern.
Source
↳ Follow the thread