MCP server for typed decisions with Jev / System One via OpenRouter or TypeSafe
View on GitHub ↗A small, dependency-light MCP server that gives any MCP client typed, probabilistic decisions from Jev, TypeSafe's System One model, through OpenRouter or TypeSafe directly.
Jev does not write text. You send it data plus typed questions and it returns one calibrated answer per question: a choice among your options, a probability for a yes/no, or a position on a scale you define, each with a confidence and the full probability distribution. It answers in a single parallel pass, fast and at a fraction of the cost of a chat model, and it cannot return an option you did not offer.
This server exposes that as five MCP tools, validates every answer the provider returns, never leaks your credential, and never retries a paid call.
- Let your agent decide instead of guess. Routing, triage, classification, claim checking and ranking become explicit, typed calls with probabilities your workflow can act on or escalate. - Cheap and fast. One paid inference call per tool invocation, regardless of how many items or questions it carries. - Provider agnostic. Use an OpenRouter key you already have, or a TypeSafe key. Switch with one environment variable. - Honest outputs. Answers are validated against the question you asked; inconsistent or incomplete answers fail the call instead of being fixed up. The unchanged provider response, including usage.cost, is always returned. - Safe by construction. Credentials live only in the server environment, HTTPS is enforced, redirects are refused, request and response sizes and time are bounded.
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/freepik-company--jev-mcp.html)
Fast, cheap, probability-scored decisions for coding agents, scripts and CI. CLI + MCP server + Claude Code plugin for the OpenAI Decisions API and Jev (TypeSafe AI).
PerryLink/jevcoreTypeSafe Jev for DeepSeek Harness, the Model Context Protocol, and plain Node: typed judgments instead of prose, offline by default.
clouatre-labs/decisions-judge-mcpTyped decisions for AI agents as an MCP tool: yes/no probability (noul), choice, and score in one fast request. Backed by the TypeSafe System One model.
Naveen-chava/metabase-mcpMetabase MCP server for AI Agents. Authenticate with a browser session cookie, no API key required.
magna-nz/tallybookPrices your Claude Code and Codex sessions from the transcripts on disk, shows what each agent cost over any period, and tells you in plain English what would have been cheaper. CLI, Browser UI and MC
Smitner-Studio/facetRender agent-written Markdown as a live local page, and send the human's click back to the agent as typed data. Human-in-the-loop approval gates for coding agents.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.