Why batch questions instead of one call each?

Jev prefills state once and scores every question in one forward pass, so extra questions add almost no latency or cost. The cited document-dominated measurement puts one batched call at 12.2 times cheaper and 10 times faster than one call per question with no change in answers.

How does triage avoid leaking file contents?

Items are text you hold or paths the server reads itself; file contents go to Jev and never enter agent context, so you screen many files then open the few that matter. Reads stay below allowed roots with symlink checks, credential names refused, and no contents in errors.

Where should the key live for reliability?

The server checks TYPESAFE_API_KEY, then JEV_API_KEY, then ~/.config/typesafe/key, and the file is most reliable because some clients strip the environment. Prefer ~/.zshenv with export for variables, and keep env blocks out so empty strings do not shadow the file fallback.

What stops the model inventing an answer?

The caller owns the option set and jev_classify requires your options, so the model can pick wrong but cannot invent. Every answer is validated, mismatched choices or legends fail as malformed_response, and a default none option lets the model decline instead of forced picking.

How do I install the server?

Requires Node 20.12+ and a TypeSafe key. One marketplace add and one plugin install register the server and the agent skill together, or you can add it with claude mcp add. Do not do both, or two servers named jev register.

Which tools does it expose?

Six: jev_classify for a Choice, jev_score for a Score, jev_check for a Noul, jev_ask for several questions about one state, jev_triage for one result per item with server-side file reads, and jev_models to confirm the key.

How do confidence gates turn into actions?

Choice and Score answers carry an action of act, review or abstain from act_above and review_above, defaulting to 0.8 and 0.5, while jev_check returns yes, no or uncertain from yes_at_or_above and no_at_or_below, defaulting to 0.7 and 0.3.

CommunityJev Agents & MCP

Jev MCP — rashedInt32

MCP server exposing Jev Choice, Score, Noul, batched ask, file-reading triage, and model listing as typed tools with full probability distributions.

See highlights

Open github.com

Stars: 7
Language: TypeScript
License: MIT
Published by: Community
Author: rashedInt32
Last verified: Oct 2026

rashedInt32/jev-mcp

User repository on GitHub

View on GitHub

MCP server exposing TypeSafe Jev as typed, calibrated judgment tools: classify, score, check, batched ask. Ships as a Claude Code plugin.

Forks: 2
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

This is one of several Jev listings for the same project. The canonical entry is Jev MCP — jkudish.

About this project

An MCP server that exposes TypeSafe Jev as typed judgment tools. Jev is a System One model. It returns a typed answer and a calibrated probability distribution, never prose. These tools surface that faithfully rather than hiding it behind a label. Requires Node 20.12+ and a TypeSafe API key from console.typesafe.ai. One marketplace add, one install. This registers the server and the agent skill together, so there is no separate skill step.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification Jev for Ranking Jev for Tool calling

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

jev_ask batches all questions over one state in a single forward pass; per-item batching also applies inside triage.
jev_triage reads path items server-side under confined roots so agents see answers without loading file contents.
Keys resolve from environment then file and never appear as tool arguments, transcripts, or model context.
Selecting tools add a none option by default and gate act, review, abstain, or yes, no, uncertain on tunable thresholds.

Watch out

Requires Node 20.12 or later and a TypeSafe key; do not register direct and plugin servers together or two jev servers register. Credential files, binaries, directories, and oversize files are refused without truncation, paths are treated as untrusted, and confidence describes distribution concentration rather than correctness. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

typesafe-mod

A Claude Code function hook that sends AskUserQuestion dialogs to Jev and draws a probability per option above the dialog, so you still choose while seeing how confident the model is.

Jev Agents & MCP#community#jevbest#catalog-import

Community

BeLazy167/typesafe-modTypeScript

typesafe-demo-mcp

A one-commit TypeScript MCP server that exposes three TypeSafe System One judgments — noul, choice and score — as agent tools, with src, tests, docs, scripts and a knip config. No README, zero stars, no topics.

Jev Agents & MCP#community#jevbest#catalog-import

Community

bestagentkits/typesafe-demo-mcpTypeScript

pi-jev-skill-picker

Pi extension that strips the full Agent Skills catalog from every request and loads only the skills a Jev ranking call judges relevant.

Jev Agents & MCP#community#jevbest#catalog-import

Community

safzanpirani/pi-jev-skill-pickerTypeScript

jevgrep

102

Behavioral code search that asks Jev to score repository fragments for a plain-language question and returns exact excerpts with paths and lines.

Jev Agents & MCP#community#jevbest#catalog-import

Community

nassim-arifette/jevgrepTypeScript

jev-judge-mcp

MCP server over stdio that gives coding agents fourteen Jev judgment tools, turning evidence plus an enumerable question into auto, review, or escalate actions from calibrated probabilities.

Jev Agents & MCP#community#jevbest#catalog-import

Community

PyModel/jev-judge-mcpPython

JevLoop

Runnable agent harness that moves pick, score and yes-no loop choices to Jev or Laya in milliseconds while reserving the language model for writing, compiling DECISION.md into executable questions.

Jev Agents & MCP#community#jevbest#catalog-import

Community

zjunlp/JevLoopTypeScript

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai