NORP Agent

Skip ▶

Skip Setup?

You can configure your API key and settings later from the Settings panel.

⚡

Welcome to NORP Agent

Your autonomous AI coding companion. Let's get you set up in just a few steps.

🌐 Interface Language / Language

🔑

API Configuration

Enter your API key and base URL to connect to the AI service.

API Base URL API Key *

🧠

Model & Workspace

Pick the model that powers your agent, and choose the workspace it operates in.

Model Project Root

🛡️

Behavior & Safety

Decide how the agent behaves and how cautious it is with your files.

Enable Web Search

Allow the Agent to search the web for real-time information.

Confirm Before Write/Delete

Prompt a confirmation dialog before writing or deleting files.

⚡

Performance & Limits

Tune how long the agent may run and how much work it can queue up.

Event Queue Limit Max Steps Task Timeout (seconds, 0 = disable)

🚀

Ready to Go

Review your settings below, then start using the agent.

API Base URL-

API Key********

Model-

Project Root-

Web SearchOff

Confirm Write/DeleteOn

Queue Limit200

Max Steps128

Task Timeout0s (disabled)

⛔ 安全系统已关闭 — 智能体行为不再受保护

Enable Safety System

Blocks dangerous commands (e.g. rm -rf /, format C:), UAC elevation, and path traversal.

✅ Safety system enabled — dangerous operations will be blocked

启用越狱/注入检测

检测并拦截试图绕过安全约束的提示词注入攻击（DAN、角色覆写、系统提示词泄露等）。关闭后所有消息将直接放行。

检测到攻击时的操作：

✅ 越狱防护已启用 — 恶意注入将被拦截

提示词来源

提示词内容

启用视觉 API

开启后，上传的图片/视频将交给视觉层处理。多模态视觉模型由开发者自行接入。

视觉外挂（vision_click / vision_type 等 L2 操作）的安全阈值。L2 操作必须经审批弹窗用户显式确认。

用户空闲判定阈值（秒）

delegate 让渡时长（秒）

熔断冷却（秒）

让渡范围

让渡最高风险级

连续失败停手次数

空闲后允许 Agent 继续操作

默认关闭：用户离开即锁死，只保留 L0 只读。开启后用户离开时 Agent 仍可操作电脑（风险自担）。

启用 capture_worker 驻留捕获（动作-验证高频循环更快）

默认关闭（每次冷启动捕获进程，最稳）。开启后捕获进程常驻复用，点击/输入后的重捕获验证延迟更低，但会持续占用一个 D3D11 捕获会话。

🔌 插件控制面板

🛠️ Agent 调试面板

This means:
• No AST code audit before loading
• No import restrictions
• No permission checks
• Any plugin code can execute freely

Only the jailbreak guard and NORP Safe system will still protect you.

With NORP Safety OFF:

• ALL dangerous commands execute freely (rm -rf /, format C:, del /F /S, etc.)
• UAC elevation / admin privilege escalation is allowed
• Path traversal outside the project workspace is permitted
• Any destructive operation on your system becomes possible

⚠️ There will be NO remaining protection layer. The Agent can do ANYTHING.

Skip Setup?

Welcome to NORP Agent

API Configuration

Model & Workspace

Behavior & Safety

Performance & Limits

Ready to Go

NORP Agent

🤖 Agent Question

Confirm Operation

Settings

📁 Project

🛡️ 原生工具确认

💾 Memory (Conversation History)

🔌 Plugins

🛡️ NORP Safety System

🔒 越狱/注入防护

🩺 运行时健康检查

📝 自定义系统提示词

👁️ 视觉 API（开发者接口）

🛡️ 视觉操作安全（外挂裁决器）

Change API Key

About

🔌 插件控制面板

Plugin Directories

已加载插件

🔒 Plugin Security

🧩 进程隔离

🔏 签名 / 来源校验

🌐 网络策略（SSRF 防护）

🛡️ 插件工具调用审批

📨 消息中心

🛠️ Agent 调试面板

⚠️ Security Warning

⚠️ CRITICAL SAFETY WARNING

🚀 Welcome to NORP Agent