jev-mcp (blakestone-x)
Python MCP server wrapping Jev as classify, score, check, match and screen tools with confidence on every answer.
Overview
This jev-mcp is a Python MCP server that exposes TypeSafe's Jev as typed tools: classify, score, check (yes/no), match and screen. Every answer comes in a stable envelope with the choice, confidence, per-label probabilities, a band such as act, token usage, estimated cost and latency. The README describes Jev as giving typed judgments over a closed answer set in about 130 ms.
It targets any MCP host; install instructions cover Claude Code, Codex and Cursor using uvx pinned to a tagged release. Registration scripts for Bash and PowerShell write the key into client configs, and the README is candid about where the plaintext key ends up. The worked example also warns that a list of texts in one state gets one combined label, so per-item labels need one request per item.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which category best fits this state? | account_access · billing · delivery · bug · feedback · other |
| Score | Rate the input on a defined scale(inferred) | — |
| Yes / no | Check whether a stated condition holds(inferred) | yes · no |
How it works
- Set TYPESAFE_API_KEY in the MCP client's environment
- Register the server with uvx at a pinned tag
- The agent calls classify, score, check, match or screen
- Use the returned band, confidence and probabilities to act or defer
Strengths
- Stable result envelope including cost and latency
- Pinned release tags for reproducible installs
- Clear guidance on key handling
Caveats
- Registration scripts write the key in plaintext to client configs
- Combined states get a single label, which can surprise users
- Requires a TypeSafe API key
Getting started
uvx --from git+https://github.com/blakestone-x/jev-mcp@v0.2.1 jev-mcp (with TYPESAFE_API_KEY set in the client environment)At a glance
- Type
- MCP server
- Status
- beta
- Language
- Python
- Runtime
- MCP server (uvx)
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
MCP ecosystem: MCP server exposing Jev classify, score, check, match, and screen as tools for any agent, with confidence on every answer.
yibie/awesome-jevPython MCP server: classify, score, check, match, and screen tools.
AnotiaWang/awesome-decision-modelsWraps Jev classification, scoring, yes/no judgments and candidate matching as MCP tools.
logicrw/awesome-jev-projectsMCP server exposing classify, score, check, match, and screen tools.
cobanov/awesome-jevMCP server for TypeSafe Jev: typed classify, score, check, match and screen for any agent, with confidence on every answer.
hellogumbo/awesome-jev
Frequently asked questions
Which tools does it expose?
classify, score, check, match and screen, each returning typed answers with confidence.
How fast is it?
The README describes Jev answers in about 130 ms; the example response reports 130 ms latency.
How do I get one label per item?
Send one request per item; a list in a single state receives one combined label.
Related projects
Unofficial hosted or local Jev MCP server giving agents a quick second opinion: it picks yes/no, scale or choice and returns probabilities.
MCP serverbetaTypeScriptSwift library and CLI for Jev with enum-typed Choice, Noul and Score questions, plus an agent skill for coding agents.
CLIproductionSwiftPython decision gateway and stdio MCP server exposing Jev via OpenRouter's alpha Decisions endpoint to Claude Code, Codex and Cursor.
MCP serverbetaPythonAgent skill and CLI that converts plain requirements, LLM prompts or prompt-calling code into a typed Jev request and runnable script.
Agent skillbetaPythonGo MCP server giving any client typed Jev decisions via OpenRouter or TypeSafe, with validated answers and no paid retries.
MCP serverproductionGoGo CLI and single-binary MCP server giving Claude Code, Codex, Hermes and pi an evaluate tool backed by Jev, d1, CLM or Laya.
MCP serverbetaGo