OpenRouter launches Armor beta to guard agent tool calls with Jev model

On October 16, OpenRouter announced the beta launch of OpenRouter Armor, a security tool designed to block out-of-policy tool calls before an agent executes them. Powered by typesafeai's Jev model, the tool reportedly outperforms Claude Code's auto classifier at a fraction of the cost.

OpenRouter noted that tool calls allow agents to interact with the outside world, but regex-based filtering is often insufficient to detect unwanted tool calls. Armor aims to strengthen protection by evaluating calls through the Jev model before execution.

OpenRouter is a platform that aggregates APIs from multiple AI model providers. The beta release gives developers early access to test the tool and provide feedback for future improvements.