What it is
An MCP server for TypeSafe's Jev, a System One model that returns typed judgments (noul, choice, score) with probabilities instead of generated text. It helps developers prototype, evaluate, and apply questions to many items.
How it uses Jev
Jev answers typed questions against a state. jev_ask prototypes a question on one state; jev_map runs a question set over many items, returning compact rows; jev_eval measures variants against labeled examples, providing accuracy, calibration, threshold sweep, and worst misses.
Primitives:noulchoicescore
Technique worth stealing
Use jev_eval to pick thresholds from a table, not intuition, and jev_map to keep bulk out of context.
Try it
claude mcp add jev -e TYPESAFE_API_KEY=sk-... -- npx -y jev-eval-mcp
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.94
- Shows a System One patternunclear0.60
- Handles uncertaintyno0.09
- Measuredno0.05
- Runnableyes0.91
- Worth recommendingyes0.65
- Model replicano0.04
- Problem scopescore on a 0–2 scale1.02
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.