What does pi-reply-guard do?

It is a Pi extension that checks each agent reply against a list of skills. A classifier model scores adherence from 0 to 1, and if a reply does not adhere to the skill, the agent is prompted to edit its reply until it does.

How does pi-reply-guard use Jev?

Jev is the default classifier model. Per the README, a classifier model (Jev by default) determines if the reply adheres to the skill on a scale of 0-1, and replies below the threshold are sent back to the agent for editing.

How do I install pi-reply-guard?

Run pi install npm:pi-reply-guard, or install from GitHub with pi install git:github.com/amalavet/pi-reply-guard. It ships from npm as the pi-reply-guard package and is also listed as a Pi package.

How is pi-reply-guard configured?

Run /reply-guard in Pi to open the settings menu, or choose Edit JSON for the full configuration. Each change saves to ~/.pi/agent/reply-guard.json and applies to the next reply, per the README.

Do I need a Jev API key to use pi-reply-guard?

The digest shows Jev as the default classifier, but key setup is not documented. Configuration is a JSON file with per-skill thresholds, such as 0.3 in the README example, edited directly or through the settings menu.

What are the limits of the digest for this project?

The README excerpt cuts off midway through the configuration example, so the complete settings list, the default thresholds, latency, and any cost per checked reply are not documented as of 2026-10-17.

Who built pi-reply-guard and under what license?

The GitHub user amalavet publishes it under the MIT license as an npm package and a Pi package. The repository was created 2026-10-07 and last pushed 2026-10-16, with 4 stars when the sweep ran.

Jev directory

CommunityJev Agents & MCP

Pi Reply Guard

A Pi extension that checks each agent reply against a list of skills: a classifier model, Jev by default, scores adherence from 0 to 1, and when a reply does not adhere the agent is prompted to edit it until it does.

See highlights Open github.com

Stars: 4
Language: TypeScript
License: MIT
Reading time: 2 min
Published by: Community
Author: amalavet
Last verified: Oct 2026

amalavet/pi-reply-guard

User repository on GitHub

View on GitHub

Checks each agent reply against a list of skills. A classifier model determines if the reply adheres to the skill on a scale of 0-1. If a reply does not adhere to the skill, the agent is prompted to edit their reply to adhere to the skill.

Forks: 0
Last pushed: Updated Oct 2026

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

Tagscommunity github pi agents

Works forJev for Decision gating Jev for Verification

Used inJev for Software & DevTools

Languages: TypeScript
Platforms: CLI
Capabilities: Open source

Highlights

Distributed on npm as pi-reply-guard and as a Pi package; install with pi install npm:pi-reply-guard.
Jev is the default classifier that scores how well a reply adheres to each configured skill.
Replies below the configured skill threshold are sent back for the agent to edit before they stand.
Settings live in ~/.pi/agent/reply-guard.json and are edited through the /reply-guard menu in Pi.
Skill thresholds are per-skill numbers, for example 0.3 in the config example shown in the README.

Quickstart

bash

pi install npm:pi-reply-guard
/reply-guard

Watch out

README reviewed 2026-10-17; the digest cuts the configuration JSON mid-example, so the full skill list, the default thresholds and how a Jev credential is supplied are not documented here.

Questions about this entry

Short answers drawn from the linked project or page.

From the Jev guides

Original write-ups on what this entry is for.

MCP tool routing: Jev in front of the tool call

Oct 16, 2026

Tool menus can outgrow the context that reads them. The catalog's Jev-based routers and gates narrow the list: benchmarked tool search, per-tool scoring, fail-closed approvals, and WebMCP in the browser.

Jev Agents & MCP#mcp#tool-routing#agents

Read article

Approval policies for coding agents: hooks, guards, and gateways