DefenseClaw is Cisco AI Defense’s open-source security governance layer for AI coding agents and agentic runtimes, with 887 stars on GitHub.
DefenseClaw provides security governance for agentic AI. As coding agents and autonomous runtimes gain the ability to call tools, run skills and reach external systems, security teams need a checkpoint: what capabilities is this agent about to use, what traffic is it generating, and can we prove later what it did? DefenseClaw answers those needs by scanning capabilities before use, inspecting runtime traffic, and exporting durable audit evidence. It combines a Python operator CLI, a Go gateway and connector hooks into one system that sits between agents and the tools they touch. The project comes from Cisco’s AI Defense team and is under very active development in 2026. That pre-use scanning model means risky capabilities can be flagged for review before an agent ever invokes them, instead of being discovered in logs after an incident. Connector hooks let existing agent frameworks plug into the gateway without being rewritten, which lowers the cost of adopting governance for teams that already run agents in production.
DefenseClaw is free and open source under the Apache-2.0 license. You deploy it within your own environment; there is no per-seat fee for the open-source project.
DefenseClaw is for security and platform teams rolling out AI coding agents or agentic runtimes across an organisation. If your CISO asks how agent tool use is vetted and audited, DefenseClaw gives you scanning, inspection and evidence export instead of a shrug.