OpenRouter Launches Armor Beta to Guard Agent Tool Calls

Key Info

OpenRouter launched Armor in beta, a plugin that uses Typesafe AI's Jev model to block out-of-policy tool calls before an agent executes them. OpenRouter says it outperforms Claude Code's auto classifier at a fraction of the cost.

Highlights

• Armor targets unwanted tool calls, since regex-based filtering is often insufficient.
• It relies on Typesafe AI's Jev model to guard calls before execution.
• On an internal test set, Jev 1.13 had the highest accuracy and lowest cost versus other Decisions models and Claude Code's use of Sonnet 5.
• Any agent doing inference through OpenRouter can opt in by adding it as a plugin with custom rules: [{ id: "armor" }].
• Docs are available for setup details.