Co-Pilot / 辅助式
更新于 25 days ago

snow-flow

Ggroeimetai
0.0k
groeimetai/snow-flow
84
Agent 评分

💡 摘要

Snow-Flow是一个开源的ServiceNow AI开发平台,能够使用自然语言轻松与其实例进行交互。

🎯 适合人群

ServiceNow开发人员AI集成专家IT服务管理团队DevOps工程师技术项目经理

🤖 AI 吐槽:README中提到的潜在风险包括OAuth凭证暴露和依赖性漏洞。为了减轻风险,确保安全存储API密钥并定期更新依赖项。

安全分析中风险

风险:Medium。建议检查:是否执行 shell/命令行指令;是否发起外网请求(SSRF/数据外发);API Key/Token 的获取、存储与泄露风险;依赖锁定与供应链风险。以最小权限运行,并在生产环境启用前审计代码与依赖。


Quick Start

# Install npm install -g snow-flow # Start the TUI snow-flow # Authenticate (in the TUI) /auth

That's it. Snow-Flow auto-initializes on first run and guides you through authentication with ServiceNow and your preferred LLM provider.


What is Snow-Flow?

Snow-Flow is an open-source AI development platform for ServiceNow. It uses the Model Context Protocol (MCP) to give AI assistants direct access to your ServiceNow instance through 400+ specialized tools.

Talk to your ServiceNow instance in plain English:

> Create a dashboard widget showing P1 incidents with SLA countdown timers

> Build a business rule to auto-assign incidents based on category

> Show me all incidents assigned to my team from the last week

> Create a knowledge article explaining the password reset process

Snow-Flow handles everything: Update Set creation, ES5 validation, widget coherence checks, and deployment.


Why Snow-Flow?

vs ServiceNow Build Agent

| Feature | Snow-Flow | ServiceNow Build Agent | |---------|-----------|------------------------| | Cost | Free (open source) | $150K-$4.5M implementation | | AI Model | 75+ providers - Claude, GPT, Gemini, DeepSeek, Ollama | NowLLM only | | Development | Any IDE - VS Code, Cursor, terminal, JetBrains | Browser-based Studio | | Integrations | Jira, Azure DevOps, Confluence, GitHub | ServiceNow only | | Source Code | 100% open source | Proprietary | | Self-hosted | Yes | No | | Architecture | Multi-agent orchestration | Single agent |

Key Features

  • 379+ MCP Tools - Complete ServiceNow coverage across 15 categories
  • Any LLM Provider - Claude, GPT, Gemini, DeepSeek, Llama, or free with Ollama
  • Built-in TUI - Interactive terminal interface with keyboard shortcuts
  • ES5 Validation - Catches Rhino engine errors before deployment
  • Update Set Management - Automatic change tracking
  • Widget Coherence - Validates HTML/Client/Server communication
  • Local Development - Pull artifacts to local files, edit with your IDE, push back
  • Multi-IDE Support - Claude Desktop, Cursor, Windsurf, Continue.dev

How It Works

┌─────────────────────────────────────────────────────────────────┐
│  Snow-Flow TUI                                                  │
│  ─────────────────────────────────────────────────────────────  │
│                                                                 │
│  You: Create an incident dashboard widget with priority charts  │
│                                                                 │
│  ┌─ Snow-Flow ─────────────────────────────────────────────┐    │
│  │ 1. Creating Update Set: "Feature: Incident Dashboard"   │    │
│  │ 2. Building widget template (HTML)...                   │    │
│  │ 3. Writing server script (ES5 validated)...             │    │
│  │ 4. Writing client controller...                         │    │
│  │ 5. Validating widget coherence...                       │    │
│  │ 6. Deploying to dev12345.service-now.com...             │    │
│  │ ✓ Widget deployed: incident_dashboard                   │    │
│  │                                                         │    │
│  │ Preview: https://dev12345.service-now.com/sp?id=...     │    │
│  └─────────────────────────────────────────────────────────┘    │
│                                                                 │
│  You: _                                                         │
└─────────────────────────────────────────────────────────────────┘

MCP Tools Overview

Snow-Flow provides 379 tools across 15 categories:

| Category | Tools | Read | Write | Description | |----------|-------|------|-------|-------------| | Platform Development | 78 | 25 | 53 | Script includes, client scripts, UI development | | Automation | 57 | 24 | 33 | Script execution, scheduled jobs, events | | Advanced AI/ML | 52 | 44 | 8 | Classification, anomaly detection, predictions | | ITSM | 45 | 17 | 28 | Incident, change, problem management | | Integration | 33 | 13 | 20 | REST messages, transform maps, import sets | | Core Operations | 30 | 14 | 16 | CRUD operations, queries, bulk operations | | UI Frameworks | 19 | 4 | 15 | Portal widgets, UI pages, forms | | Security | 18 | 9 | 9 | ACLs, policies, compliance | | CMDB | 14 | 10 | 4 | CI management, relationships, discovery | | Reporting | 10 | 3 | 7 | Reports, dashboards, KPIs | | UI Builder | 9 | 2 | 7 | Now Experience components | | Asset Management | 8 | 1 | 7 | Asset lifecycle, contracts | | Performance Analytics | 3 | 2 | 1 | PA indicators, breakdowns | | ML Analytics | 2 | 2 | 0 | Native ML features | | Workspace | 1 | 0 | 1 | Agent Workspace config |

Browse all tools: snow-flow.dev/en/mcp-reference


Setup Guide

1. Install Snow-Flow

npm install -g snow-flow

2. Start the TUI

snow-flow

On first run, Snow-Flow automatically creates:

  • .mcp.json - MCP server configuration
  • CLAUDE.md / AGENTS.md - AI instructions

3. Authenticate

In the TUI, type /auth to authenticate:

  1. LLM Provider - Claude, GPT, Gemini, Ollama, etc.
  2. ServiceNow - OAuth 2.0 (requires OAuth app in ServiceNow)

Detailed OAuth setup: snow-flow.dev/en/oauth-setup


ServiceNow OAuth Setup

Create an OAuth application in your ServiceNow instance:

  1. Navigate to System OAuth → Application Registry
  2. Click New → Create an OAuth API endpoint for external clients
  3. Configure:
    • Name: Snow-Flow
    • Redirect URL: http://localhost:3005/callback
  4. Copy Client ID and Client Secret

In Snow-Flow TUI:

/auth

Follow the prompts to enter your instance URL and OAuth credentials.


LLM Provider Configuration

Snow-Flow supports 75+ AI providers. Configure via environment variables or /auth:

Claude (Recommended)

export ANTHROPIC_API_KEY=sk-ant-api03-...

Or use Claude Pro/Max subscription via browser auth.

OpenAI

export OPENAI_API_KEY=sk-...

Google Gemini

export GOOGLE_API_KEY=AIza...

100% Free - Local Ollama

# Install: https://ollama.com ollama pull llama3.3 export OLLAMA_BASE_URL=http://localhost:11434

TUI Keyboard Shortcuts

| Shortcut | Action | |----------|--------| | Ctrl+X | Leader key (prefix for commands) | | Tab | Switch to next agent | | F2 | Cycle through recent models | | Enter | Submit message | | Shift+Enter | Insert newline | | Esc | Interrupt operation | | Ctrl+X H | Show help | | Ctrl+X N | New session | | Ctrl+X M | List models |

Full keyboard reference: snow-flow.dev/en/tui-guide


CLI Commands

snow-flow # Start TUI snow-flow auth login # Authenticate snow-flow auth list # Show auth status snow-flow sessions list # List sessions snow-flow config show # Show configuration snow-flow models # List available models snow-flow run "prompt" # Run single prompt snow-flow stats # Show usage stats

Full CLI reference: snow-flow.dev/en/cli-reference


Use With Any AI IDE

Snow-Flow works with any MCP-compatible tool:

| Tool | Setup | |------|-------| | Snow-Flow TUI | Built-in, just run snow-flow | | Claude Desktop | Copy .mcp.json to ~/Library/Application Support/Claude/ | | Cursor | Copy to .cursor/mcp.json | | Windsurf | Copy to mcp_config.json | | Continue.dev | Add to ~/.continue/config.json |

Reducing Token Usage with External Clients

When using Snow-Flow MCP with external clients like Claude Code, all 235+ tools are loaded at once (~71k tokens). Use lazy loading to reduce this dramatically:

Option 1: Lazy Tools (Recommended) - ~2k tokens

{ "mcpServers": { "snow-flow": { "command": "npx", "args": ["snow-flow", "mcp"], "env": { "SNOW_LAZY_TOOLS": "true" } } } }

With lazy tools, only 2 meta-tools are exposed:

五维分析
清晰度8/10
创新性8/10
实用性9/10
完整性8/10
可维护性9/10
优缺点分析

优点

  • 完全开源且免费使用。
  • 支持多个AI提供商以增加灵活性。
  • 丰富的工具集覆盖全面的ServiceNow功能。

缺点

  • 初始设置可能需要技术知识。
  • 高级功能的文档有限。
  • 依赖外部LLM提供商以实现完整功能。

相关技能

pytorch

S
toolCode Lib / 代码库
92/ 100

“它是深度学习的瑞士军刀,但祝你好运能从47种安装方法里找到那个不会搞崩你系统的那一个。”

agno

S
toolCode Lib / 代码库
90/ 100

“它承诺成为智能体领域的Kubernetes,但得看开发者有没有耐心学习又一个编排层。”

nuxt-skills

S
toolCo-Pilot / 辅助式
90/ 100

“这本质上是一份组织良好的小抄,能把你的 AI 助手变成一只 Nuxt 框架的复读机。”

免责声明:本内容来源于 GitHub 开源项目,仅供展示和评分分析使用。

版权归原作者所有 groeimetai.