shieldclaw-mcp

ShieldClaw is an independent governance layer for AI agents. It is described as AI agent security infrastructure and is meant to make agents such as Claude, GPT, Kimi, and OpenClaw safe to deploy in production. The product is presented as open source on GitHub, and the site says that every plan includes a 14-day free trial with no credit card required.

The suite is divided into five layers. Monitor provides real-time monitoring and alerts for OpenClaw deployments, with breach alerts, a security dashboard, and compliance reports. Hunt adds weekly pentesting for OpenClaw deployments, covering 50+ attack scenarios and AI-generated fixes with 1-click remediation. Firewall sits between agents and tools as a proxy, inspects every request, blocks attacks in under 2 milliseconds, and includes an intent analysis engine, prompt injection blocking, and custom rules. Frontier Translation converts OpenClaw event logs into enterprise-grade audit trails and supports SOC2, HIPAA, GDPR, and ISO27001, with immutable audit trails and PDF compliance reports.

Shield-MCP is the standalone permission layer in the stack. Every AI agent, including Claude, GPT, Kimi, or any other model, must connect to Shield-MCP before accessing any tool, database, or system. The page illustrates grant, deny, and approval outcomes for different requests, and says the system technically restricts what agents can physically do. Delivery is subscription-based, with listed pricing for Monitor at $79 per month, Hunt at $199 per month, Firewall at $299 per month, Frontier at $399 per month, Firewall plus Frontier at $599 per month with Hunt included, and Shield-MCP custom enterprise pricing starting from $1,499 per month. The enterprise tier includes the full product suite, a Shield-MCP server, support for all models, a white-label option, and dedicated support.