Cloud Firewall releases the Agent Firewall feature
Zero intrusion into the agent runtime environment, with deep inspection, access control, and behavior audit of all agent outbound traffic (access to LLM APIs, external services, and MCP Server).
Content
Applicable customers: Enterprise users deploying autonomous agents such as OpenClaw on the cloud need to audit and control the network behavior of agents autonomously invoking third-party Large Language Model (LLM) APIs.
New Feature/Specification:
1. Supports one-click enablement of Agent Firewall protection for the Internet egress of the agent network environment, and automatically detects the type and quantity of associated agents.
2. Supports agent behavior control features including: access control (model limitations, network exposure protection), outlier detection (skills&tools invocation anomaly alerting, skill file threat detection), and data breach detection (AK/API key leak detection, personal information leak detection, and sensitive file exfiltration detection).
3. Supports log audit of agent behavior control security management events and agent access to LLM, MCP Server, and external services.