Show HN: Jev Decision Layer: Save Frontier Tokens on Closed Decisions (github.com)

Qualixar has unveiled the Jev Decision Layer, an open-source MCP decision layer designed to enhance the efficiency of AI agents by streamlining the decision-making process through fast, typed choices. This layer allows developers to route various tasks—such as tool selection and skill choice—through the TypeSafe Jev platform, which boasts impressive benchmarks: Jev performs 193.6 times faster and is 444.6 times more cost-effective than conventional large language models (LLMs) for these workflows. With version 1.0.8, the release includes 20 MCP tools, 38 bounded decision recipes, and support for multiple host adapters, including Codex and Antigravity. The significance of this development lies in its ability to create a more effective workflow for AI agents by facilitating decision routes and local policy checks without unnecessary interactions with chat models. Furthermore, the layer includes a local receipt system that records decisions while the host maintains control over execution, enhancing accountability in operations. This approach not only improves the speed and cost of decision-making but also provides a structured framework for integrating different AI models, as organizations navigate the complexities of automated workflows while maintaining oversight and accuracy. The layer is currently supported on macOS, with Linux and Windows capabilities still under development.