OpenRouter Launches Armor to Block Unauthorized AI Agent Tool Calls

AI5 days ago

A new beta tool from OpenRouter uses decision models to intercept unauthorized actions by AI agents before they interact with external systems. Called Armor, the product operates as an "auto" permissions mode compatible with any AI model or harness, utilizing a specialized classifier to replace traditional regex-based filtering that the company describes as insufficient for detecting complex policy violations.

The system is powered by Typesafe AI's Jev model. Internal testing by OpenRouter found that Jev 1.13 achieved higher accuracy and lower costs than other available decision models and the classifier employed by Claude Code's Sonnet 5. The beta release aims to provide a scalable guardrail for autonomous agents executing tool calls in real-world environments.

Mapped to your portfolio

Possible scenarios

40%Armor Gains Adoption+0.3%
25%Incumbents Absorb Controls+0.2%
20%Attackers Bypass Armor-0.3%

Each figure is the modeled move of the S&P 500 if the outcome occurs. For information only. Not investment advice.

Looking for more context or analysis?

Tilt chat provides reliable, tailored insights, alerts, and automations.

OpenRouter Launches Armor to Block Unauthorized AI Agent Tool Calls

In this story

OpenRouter Typesafe AI

Portfolio

SPY
State Street SPDR S&P 500 ETF Trust
100%

Portfolio

SPY
State Street SPDR S&P 500 ETF Trust
100%

Mapped to your portfolio

Possible scenarios

40%Armor Gains Adoption+0.3%
25%Incumbents Absorb Controls+0.2%
20%Attackers Bypass Armor-0.3%

Each figure is the modeled move of the S&P 500 if the outcome occurs. For information only. Not investment advice.