ServiceNow AI工具

Getting Started

Option A — Interactive Setup Wizard (Recommended)

```bash

Install globally (Node.js 20+ required)

npm install -g nowaikit

Option D — Manual Setup

git clone https://github.com/aartiq/nowaikit.git && cd nowaikit
npm install && npm run build
cp .env.example .env   # fill in your ServiceNow credentials

Then point your AI client at dist/server.js — see Supported AI Clients below.

No ServiceNow instance? Get a free Personal Developer Instance at developer.servicenow.com — ready in minutes.

Full installation guide → docs/INSTALLATION.md

---

Client Setup Guides

Step-by-step setup for every major AI client — Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Zed, GitHub Copilot, Continue.dev, Cline, JetBrains, Amazon Q, Google AI Studio, ChatGPT, Grok, Ollama, and more.

Full guide → docs/CLIENT_SETUP.md

For quick setup snippets, see the [Supported AI Clients](#supported-ai-clients) section below.

Install Claude Code

npm install -g @anthropic-ai/claude-code

Zero-config setup & desktop experience

• npx nowaikit setup — interactive wizard detects and writes config for every AI client automatically. No JSON editing.
• nowaikit Desktop — cross-platform Electron app (macOS, Windows, Linux) with 8-step visual wizard, dashboard, tool browser, and audit log. No separate server install needed.
• Web dashboard — served at http://localhost:3100 by the HTTP server. Browse tools, view health, tail audit logs.

App builder integrations

• HTTP API server (npm run serve) — REST proxy for Lovable, Bolt, v0, Replit apps. Keeps credentials server-side.
• Smithery registry — smithery install nowaikit one-command install.

Full changelog

Example Interactions

Once connected, ask your AI assistant in plain language:

ITSM & Change Management:

Show me all open P1 incidents assigned to the Network Operations group.

Create a normal change request for deploying the new API gateway — implementation planned for Saturday midnight.

What CMDB CIs does the ERP application depend on?

Scripting & Development:

List all client scripts on the incident table and show me the ones that fire on form load.

Create a UI action button "Escalate to L3" on the incident form that assigns the ticket to the L3-Support group.

Show me all ACL rules for the change_request table that restrict the "delete" operation.

Service Portal & UI Builder:

List all widgets in the Service Portal that contain "catalog" in their name.

Get the full source code of the "Stock Ticker" widget so I can update its server script.

Create a new portal widget called "My Approvals Widget" with a simple Angular template that lists pending approvals.

Integrations & Events:

List all REST Message definitions that connect to external APIs.

Show me all transform maps that target the incident table.

Fire the custom event "myapp.ticket.escalated" on incident INC0012345.

Notifications & Attachments:

List all email notifications that trigger on the incident table when a comment is added.

Upload a screenshot of the error (base64) as an attachment to incident INC0012345.

Show me all failed email log entries from the last 24 hours.

Performance Analytics & Data Quality:

Get the current scorecard for the "Mean Time to Resolve" PA indicator with a 30-day trend.

Check the data completeness of the incident table — how many incidents are missing assignment_group or category?

Compare record counts across incident, change_request, and problem tables.

ATF, Reporting & Scheduled Jobs:

Run the Regression Test Suite and show me any failures with ATF Failure Insight details.

Summarise the last 30 days of incident trends by category.

Create a scheduled job that runs daily at 3am to archive closed incidents older than 90 days.

For 120+ real-world examples with inputs, outputs, and advanced workflows, see EXAMPLES.md.

---

Run the wizard — detects your AI clients and writes config automatically

npx nowaikit setup

The wizard will:
1. Ask for your ServiceNow instance URL + credentials
2. Test the connection
3. Let you pick a tool package and permission level
4. Detect Claude Desktop, Cursor, VS Code, Windsurf, Continue.dev, Claude Code on your machine
5. Write the config directly — no copy-paste, no manual JSON editing

Option B — Web Dashboard (one command)

npx nowaikit web

Opens the NowAIKit dashboard at http://localhost:4175 — includes instance management, settings, and an audit log viewer.

```bash

Option C — Desktop App

Download the native desktop app (macOS, Windows, Linux) from GitHub Releases. Includes a visual setup wizard, tool browser, and audit log viewer.

Advanced Configuration

Use with Responses API tool type "mcp" or via function calling

response = client.responses.create( model="gpt-4o", tools=[{"type": "mcp", "server_label": "nowaikit"}], input="Show me all open P1 incidents" )

Full guide → [docs/CLIENT_SETUP.md](docs/CLIENT_SETUP.md)
</details>

<details>
<summary><b>Google Gemini / Vertex AI</b> (API)</summary>

1. Install NowAIKit: `npm install -g nowaikit`
2. Use the Python client in `clients/gemini/`:

Full guide → [clients/gemini/SETUP.md](clients/gemini/SETUP.md)
</details>

<details>
<summary><b>Google AI Studio</b> — Gemini 2.5 Flash / Pro / Gemini 3 (MCP Preview)</summary>

Google AI Studio supports MCP servers via its agent execution environment (currently in preview).

1. Go to [aistudio.google.com](https://aistudio.google.com) and open **Build → Agent**
2. In the **Tools** panel, add an MCP Server and point it to your NowAIKit instance:

3. Alternatively, use the **Gemini API** directly with function calling by mapping NowAIKit tool definitions:

CLI & Terminal Agents

<details> <summary><b>Claude Code / Claude CLI</b> — Anthropic's official CLI</summary>

```bash

Programmatic / Agent SDK

<details> <summary><b>OpenAI Codex / Custom Python Agent</b></summary>

cd clients/codex
pip install -r requirements.txt
cp .env.basic.example .env  # fill in your credentials
python servicenow_openai_client.py

Full guide → clients/codex/SETUP.md </details>

<details> <summary><b>Anthropic Agent SDK (Claude API)</b></summary>

```python import anthropic, subprocess, json

Then use with the Anthropic MCP client SDK

```

See Anthropic MCP Python SDK for full integration. </details>

---

Quick Reference

---

Module Coverage

Domain modules covering the full ServiceNow platform:

---

Role-Based Tool Packages

Set MCP_TOOL_PACKAGE to expose only the tools relevant to each persona:

---

IDE Extensions

<details> <summary><b>VS Code</b> — Native MCP (v1.99+, no subscription required)</summary>

VS Code 1.99 and later includes built-in MCP support — no extension or subscription required.

1. Install VS Code 1.99+
2. Create .vscode/mcp.json in your workspace (or edit User settings):

{
  "servers": {
    "nowaikit": {
      "type": "stdio",
      "command": "node",
      "args": ["/absolute/path/to/nowaikit/dist/server.js"],
      "env": {
        "SERVICENOW_INSTANCE_URL": "https://yourinstance.service-now.com",
        "SERVICENOW_AUTH_METHOD": "basic",
        "SERVICENOW_BASIC_USERNAME": "admin",
        "SERVICENOW_BASIC_PASSWORD": "your_password",
        "WRITE_ENABLED": "true",
        "SCRIPTING_ENABLED": "true"
      }
    }
  }
}

1. Open the Command Palette (Cmd/Ctrl+Shift+P) → MCP: List Servers to verify the connection
2. Open Copilot Chat (or any AI assistant in VS Code) and use @nowaikit or just ask naturally

Tip: Add .vscode/mcp.json to .gitignore if it contains credentials, or use environment variables from a .env file.

Full guide → clients/vscode/SETUP.md </details>

<details> <summary><b>VS Code — GitHub Copilot</b> (agent mode)</summary>

1. Install VS Code + GitHub Copilot extension
2. Create .vscode/mcp.json in your project:

{
  "servers": {
    "nowaikit": {
      "type": "stdio",
      "command": "node",
      "args": ["${workspaceFolder}/../../nowaikit/dist/server.js"],
      "env": {
        "SERVICENOW_INSTANCE_URL": "https://yourinstance.service-now.com",
        "SERVICENOW_AUTH_METHOD": "basic",
        "SERVICENOW_BASIC_USERNAME": "admin",
        "SERVICENOW_BASIC_PASSWORD": "your_password"
      }
    }
  }
}

Full guide → clients/vscode/SETUP.md </details>

<details> <summary><b>VS Code — Continue.dev</b> (open-source Copilot alternative)</summary>

1. Install Continue from VS Code Marketplace
2. Edit ~/.continue/config.json:

{
  "mcpServers": [
    {
      "name": "nowaikit",
      "command": "node",
      "args": ["/path/to/nowaikit/dist/server.js"],
      "env": {
        "SERVICENOW_INSTANCE_URL": "https://yourinstance.service-now.com",
        "SERVICENOW_AUTH_METHOD": "basic",
        "SERVICENOW_BASIC_USERNAME": "admin",
        "SERVICENOW_BASIC_PASSWORD": "your_password"
      }
    }
  ]
}

Full guide → docs/CLIENT_SETUP.md </details>

<details> <summary><b>VS Code — Cline</b> (autonomous AI agent)</summary>

1. Install Cline from VS Code Marketplace
2. Open Cline → MCP Servers → Add Server
3. Enter the path to nowaikit/dist/server.js and your environment variables

Full guide → docs/CLIENT_SETUP.md </details>

<details> <summary><b>JetBrains AI Assistant</b> (IntelliJ IDEA, PyCharm, WebStorm, etc.)</summary>

1. Install JetBrains AI Assistant plugin
2. Go to Settings → Tools → AI Assistant → MCP Servers
3. Add a new server with the path to nowaikit/dist/server.js
4. Set environment variables in the server configuration dialog

Full guide → docs/CLIENT_SETUP.md </details>

<details> <summary><b>Amazon Q Developer</b> (AWS CLI + IDE)</summary>

1. Install Amazon Q Developer extension for VS Code or IntelliJ
2. Configure MCP via ~/.aws/amazonq/mcp.json:

{
  "mcpServers": {
    "nowaikit": {
      "command": "node",
      "args": ["/path/to/nowaikit/dist/server.js"],
      "env": {
        "SERVICENOW_INSTANCE_URL": "https://yourinstance.service-now.com",
        "SERVICENOW_AUTH_METHOD": "basic",
        "SERVICENOW_BASIC_USERNAME": "admin",
        "SERVICENOW_BASIC_PASSWORD": "your_password"
      }
    }
  }
}

Full guide → docs/CLIENT_SETUP.md </details>

---

Frequently Asked Questions

Do I need to know the ServiceNow API to use this? No. For beginners, you just connect your AI and ask questions in plain English. The kit handles all API calls automatically.

Which ServiceNow versions are supported? All actively supported ServiceNow releases. The toolkit targets the latest available APIs and has been tested on the three most recent releases — Zurich, Yokohama, and Xanadu — and works on any currently supported instance.

Can I use this on a free Personal Developer Instance (PDI)? Yes. Get a free PDI at developer.servicenow.com and connect in 5 minutes.

Is it safe to use on production? Yes. The permission system is read-only by default. Write, scripting, and Now Assist capabilities must each be explicitly enabled with environment variables. Use role packages to limit the tool surface.

Can I use multiple AI providers at the same time? Yes. Each AI client gets its own MCP config pointing at the same (or different) NowAIKit instance. Run Claude Desktop and Cursor side by side against the same ServiceNow environment. In the nowaikit Desktop app you can switch between Claude, ChatGPT, and Gemini with one click inside the Chat page.

Does it support multi-instance / multiple customers? Yes. Configure any number of instances (prod, staging, dev, or multiple customer tenants) via instances.json or environment variables. Use list_instances, switch_instance, and get_current_instance tools to manage them, or pass instance: "name" to any individual tool call. See docs/MULTI_INSTANCE.md.

Is it free? Completely free to use for personal and commercial purposes. See LICENSE for details.

---