cisco-ai-defense/defenseclaw

Security Governance for Agentic AI

What it solves

DefenseClaw provides security governance for agentic AI runtimes and OpenClaw. It addresses the risk of using untrusted AI agent capabilities by allowing operators to scan capabilities before they are used, inspect traffic during runtime, and maintain durable audit evidence for compliance and security monitoring.

How it works

The system functions as an enforcement and evidence layer consisting of several integrated components:

  • Go Gateway: Acts as the primary traffic and enforcement point.
  • Python Operator CLI/TUI: Provides the interface for operators to manage the system.
  • Connector Hooks: Integrates with AI runtimes to intercept and monitor activity.
  • Policy and Scanners: Defines the rules and scanning mechanisms used to evaluate capabilities.
  • Observability Exporters: Sends security data to external monitoring tools.

Who it’s for

It is designed for security operators and administrators who need to govern the behavior of AI agents and ensure that the capabilities they execute are scanned and audited.

Highlights

  • Pre-use Scanning: Ability to scan AI agent capabilities before they are executed.
  • Runtime Inspection: Real-time monitoring of traffic within agentic AI runtimes.
  • Audit Trails: Export of durable evidence for security auditing.
  • Multi-language Architecture: Built using a combination of Go for the gateway and Python for the management CLI.

Related

  • Dispatch
    Securing the Agentic Frontier: Inside OpenClaw's Defense-in-Depth Strategy

    OpenClaw outlines its approach to securing powerful AI assistants through filesystem boundaries, network egress control, and rigorous plugin provenance.

  • Project
    canyonroad/agentsh

    A secure, policy-enforced execution gateway for AI agents that intercepts and governs file, network, process, and database activity at runtime.

  • Project
    microsoft/agent-governance-toolkit

    A governance framework for autonomous AI agents that provides deterministic policy enforcement, identity management, and audit logging to prevent unauthorized actions.

  • Project
    denoland/clawpatrol

    A security firewall for AI agents that intercepts network traffic and gates actions against HCL-defined rules to prevent unauthorized production changes.

  • Project
    luckyPipewrench/pipelock

    An open-source AI agent firewall that provides verifiable egress control by inspecting network traffic to prevent secret exfiltration and prompt injection.