What it is
A Python MCP server that exposes TypeSafe Jev's typed classify, score, check, match, and screen judgments as tools, so any MCP-capable agent can get calibrated confidence on closed-answer decisions.
How it uses Jev
Each tool sends a state plus a typed question (Choice, Score, or Noul) to Jev and returns a structured envelope with the answer, confidence, probabilities, and band. Agents call jev_classify, jev_score, jev_check, jev_match, or jev_screen; policy thresholds live in jev_mcp/questions.py.
Primitives:choicescorenoul
Technique worth stealing
Order-ensemble turns disagreement between reversed criteria orders into a cheap review signal.
Try it
Set TYPESAFE_API_KEY, then run uvx --from git+https://github.com/blakestone-x/jev-mcp@v0.2.1 jev-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-centricyes0.95
- Shows a System One patternno0.32
- Handles uncertaintyno0.05
- Measuredno0.26
- Runnableyes0.98
- Worth recommendingno0.32
- Model replicano0.04
- Problem scopescore on a 0–2 scale1.40
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.