Will Jev ever approve a command or certify my task done?

No. Results are advisory only; the hook can only force an approval prompt or block a flagged command, and any local failure leaves the harness unchanged. Permission rules and tests stay in charge.

What happens when the key, budget, or network fails mid-run?

The client returns an explicit unavailable result and the workflow carries on as if Jev were absent, so build every call site with a no-Jev fallback path from the start.

How do I install it without leaking my API key?

Run jev setup to store the key in DPAPI or the OS keychain, never in a config file; harness configs carry only variable references, and headless machines can point setup at an environment variable name.

Do I install a published release or a development branch?

The README installs the v0.3 changes from a reviewed development branch, not a published v0.3 release; MCP support needs Python 3.10+, while the core library and CLI run on 3.9+.

How much can Jev spend before the budget stops it?

A daily budget is required; before each request the shared ledger reserves the worst-case cost, so the cap holds across every harness sharing the same JEV_HOME, and a daily budget of 0 keeps requests disabled.

Does my state get redacted before it leaves the machine?

Yes. Question IDs, labels, and state are scanned for recognizable secrets and redacted before they leave the machine, and the model jev-1.13.0 and the official HTTPS endpoint are pinned with no proxy discovery or redirect following.

Which MCP tools does jev-decision expose?

jev_decide for batches of typed questions, jev_guard_command for shell risk, jev_verify_completion for goal gaps, jev_read_evidence and jev_prune_output for logs, and jev_status, which reports local config and budget with no network call.

CommunityJev Agents & MCP

Jev Decision

Portable Python client bringing typed Jev judgments (noul, choice, score) to coding-agent harnesses over MCP or shell, with a shared budget ledger plus harness and memory-system recipes.

See highlights

Open github.com

Stars: 1
Language: Python
License: MIT
Published by: Community
Author: Coding-Dev-Tools
Last verified: Oct 2026

Coding-Dev-Tools/jev-decision

User repository on GitHub

View on GitHub

Zero-dependency System 1 decision engine, calibrated guardrails, and token optimization client for Jev (TypeSafe AI)

Forks: 0
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Fast, typed, budgeted TypeSafe Jev decisions for coding-agent harnesses: Command Code, Claude Code, Codex, Cursor, Gemini CLI, Antigravity, OpenCode, and any MCP or shell-capable client.

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

Tagscommunity github

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Python

Highlights

Advice, never authority: failures return explicit unavailable, and harness permission rules and executed tests stay in charge.
jev setup stores the key in DPAPI or the OS keychain with a daily budget; the ledger reserves the worst-case cost per request.
Harness installer wires Command Code, Claude Code, Codex, Cursor, Gemini CLI, Antigravity, OpenCode and more; the shell hook is escalate-only.
MCP tools cover decide, guard_command, verify_completion, read_evidence, prune_output and status; evidence capture keeps big logs out of context.

Watch out

Unofficial harness client; judgments are advisory and thresholds stay in downstream code. 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

wald-4b

Open-weight 4B decision model on Qwen3.5-4B-Base serving a Jev-compatible /v1/systemone API, returning per-option probabilities with adjustable thinking.