What is agent-reflex?

A plugin and skill for coding-agent workflows. It guards shell commands, runs syntax checks after edit and write calls, compacts history by decision, and adapts thinking effort, all using a System One decision model as the classifier.

Which classifiers can it use?

TypeSafe Jev through the TypeSafe API, OpenRouter, the Vercel AI Gateway or OpenCode Zen, or local System One models served by llama.cpp 0.6+, Ollama 0.35+ or LM Studio 0.4.26+ at their /systemone endpoint.

Install the skill with npx skills add psygreg/agent-reflex, or bun x skills add, and ask the agent to install it, or run ./install.sh. A fresh install asks for the classifier provider, endpoint, model and an optional key.

Which coding agents are supported?

Eight: OpenCode, Pi, MiMo Code, Claude Code, Qwen Code, Codex, Antigravity and Hermes. The README calls OpenCode the reference implementation and the only agent where every feature works.

What differs between the supported agents?

The shell guardrail, post-edit checks and classify and confidence features work nearly everywhere; dynamic effort and custom history pruning only reach agents whose hook APIs can rewrite the model request, so Claude Code, Qwen Code, Codex, Antigravity and Hermes lose them.

What does dynamic thinking effort do?

It qualifies each prompt as low, medium or high complexity and sets maxTokens alongside effort changes for wired dialects covering Anthropic, GPT reasoning, GLM, MiMo, Gemini, Qwen and Kimi models; unwired models keep their defaults.

The skill briefs your agent on the plugin's capabilities and installs or updates it. Updates need no answers and can run unattended with bash install.sh </dev/null; re-running the installer leaves configured options untouched.

Jev directory

CommunityJev Agents & MCP

Agent Reflex

A plugin and skill for coding agents that guards shell commands, compacts history and adapts thinking effort using any System One decision model: TypeSafe Jev via API, OpenRouter, Vercel AI Gateway, OpenCode Zen or a self-hosted llama.cpp, Ollama or LM Studio server.

See highlights Open github.com

Stars: 2
Language: TypeScript
License: APACHE-2.0
Reading time: 2 min
Published by: Community
Author: psygreg
Last verified: Oct 2026

psygreg/agent-reflex

User repository on GitHub

View on GitHub

Plugin for coding agent workflow optimization and safeguarding using any system one decision model, from Jev to self-hosted LM Studio, Ollama or llama.cpp. The skill can be used to install or update the plugin through the agent itself.

Forks: 0
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity github typescript opencode

Works forJev for Decision gating Jev for Classification

Used inJev for Software & DevTools

Languages: TypeScript
Platforms: OpenRouter
Capabilities: Open sourceRuns locally

Highlights

Classifier can be TypeSafe Jev via API, OpenRouter, Vercel AI Gateway, OpenCode Zen or a local llama.cpp, Ollama or LM Studio server.
Guards shell commands, runs syntax checks after edit and write, and compacts history by decision.
Dynamic thinking effort qualifies each prompt low, medium or high and sets maxTokens for wired dialects.
Eight agents supported; OpenCode is the reference implementation where every feature works.
Apache-2.0; install via npx skills add psygreg/agent-reflex or ./install.sh.

Quickstart

bash

npx skills add psygreg/agent-reflex
./install.sh

Watch out

README and GitHub metadata reviewed 2026-10-17; the digest truncates the effort-dialect paragraph. The plugin's own README says full feature parity holds only on OpenCode, while other agents have documented hook-API limitations that disable some features.

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

Oct 16, 2026

How the Jev guard family turns coding-agent permissions into auto-approve, deny, or ask: hooks that enforce, guards that fail safe, and gateways that gate generation.

Jev Agents & MCP#coding-agents#guardrails#hooks

Read article

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai