When does the mod deliberately refuse to nudge and just stop?

Waiting on the user for permission, info, or a choice stops it, as does a low nudge score, a high waiting or stalled-progress score, any error, or hitting the per-turn nudge cap.

What happens to my agent run if the Jev request itself fails?

The run ends normally. A failed judgment never forces continuation, which keeps a Jev outage from turning into an infinite agent loop.

How do I rein in a nudge-happy agent without uninstalling the mod?

Raise the probability threshold, set a max-nudges cap with zero disabling it entirely, or turn on verbose mode plus debug logging to see skipped nudges.

Can this loop forever on a task the agent cannot finish?

No. The cap is eight nudges per user turn by default, the progress check blocks nudging when the last nudge achieved nothing, and an all-done answer ends the run.

How do I install the nudge mod?

Run cmd login if you are not signed in, then cmd mods add cmd-mod-jev-nudge. Add -g to install it for all projects. It uses your Command Code key via CMD_API_KEY or cmd login.

What does it ask Jev at each agent stop?

Three questions: would a nudge help right now, does the agent need you for permission, info, or a choice, and did the last nudge get anywhere, with the progress question starting from the second nudge.

How do I develop against it without spending calls?

Offline, run node scripts/fake-jev.mjs and add --mod-option jev-nudge-base-url=http://127.0.0.1:4711; it logs requests to jev-log.jsonl, and DEBUG=true logs every decision to stderr.

CommunityJev Agents & MCP

Cmd Mod Jev Nudge

Command Code mod that asks Jev at every agent stop whether a nudge would help, continuing unfinished work while leaving waits for the user alone.

See highlights

Open github.com

Stars: 17
Language: TypeScript
License: MIT
Published by: Community
Author: CommandCodeAI
Last verified: Oct 2026

CommandCodeAI/cmd-mod-jev-nudge

Organization repository on GitHub

View on GitHub

Command Code mod: nudges the agent to keep going when it stops with work left, judged by Jev

Forks: 4
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Agents stop too early. They do one step, say "next I'll do X", and hand the prompt back. This Command Code mod asks Jev at every stop: would a nudge help? Yes, and the agent keeps going. It won't nudge when the agent is waiting on you, or when the last nudge got nothing done. That's it. Uses your Command Code key (CMDAPIKEY or cmd login). If Jev fails, the run stops as normal.

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

Tagscommunity jevbest catalog-import

Works forJev for Classification Jev for Evaluation Jev for Tool calling

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Three-probability check per stop covering nudge value, waiting-on-user, and recent progress.
Continues multi-step tasks from one prompt while respecting explicit wait-for-me instructions.
Tunable threshold, cap, verbosity, endpoint, and model flags with debug logging to stderr.
Offline fake server for development that logs requests without spending live calls.

Watch out

If Jev errors, the run stops as normal rather than guessing, and verbosity plus the fake local server exist so decisions can be inspected without spending calls. 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

zevals

TypeScript evaluation utilities that test full AI agent conversations as end-to-end scenarios with binary assertions, without imposing a test runner.

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

Community

opencx-labs/zevalsTypeScript

jev-mcp

An MCP server built around measuring questions rather than asking them: jev_ask prototypes wording, jev_eval sweeps thresholds against labeled data, jev_map applies the winner over a workload.

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

Community

BYK/jev-mcpTypeScript

duet-agent

Agent harness for work that outlives one chat, with resumable turns, routed state machines, and observational memory across sessions and processes.

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

Community

dzhng/duet-agentTypeScript

skilled-agent-harness_spec-driven-loops

Assistant framework that stores specs, memory, and specialist agents as plain markdown files in your repo for Claude Code, Codex, Opencode, and similar runtimes.

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

Community