What it is
A local stdio MCP server for agents that need a decision made. Send a decision, context, and 2–64 choices; receive a recommended choice with percentages. Defaults to TypeSafe AI Jev via AI Gateway.
How it uses Jev
Jev is the default evaluation model (typesafe-ai/jev). The server sends the decision, context, and choices through AI SDK 7's experimental_evaluate; Jev returns choice probabilities, which are multiplied by 100 as provider-distribution percentages. The agent then decides what to do next.
Primitives:choice
Technique worth stealing
Two-stage decision tree: a router selects a bias profile by description, then evaluation runs under that policy; explicit per-profile tools bypass routing.
Try it
npx -y decide-mcp@0.1.0 with AI_GATEWAY_API_KEY set in the MCP client env.
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-centricyes0.66
- Shows a System One patternunclear0.53
- Handles uncertaintyno0.06
- Measuredno0.05
- Runnableyes0.98
- Worth recommendingno0.30
- Model replicano0.05
- Problem scopescore on a 0–2 scale1.37
- About Jevyes0.97
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.