Build a ServiceNow ITSM MCP Server for Agentic Incident Management & Change Control in 2026
IT teams spend 2.8 hours per incident on manual ServiceNow ticket navigation. This FastMCP server exposes incident CRUD, change request workflows, and CMDB queries to AI agents, enabling autonomous incident triage and resolution directly from Claude Desktop or Cursor IDE.
Deepak Bagada
CEO, SaaSNext
- ServiceNow MCP server reduces incident handling time from 2.8 hours to under 10 minutes per ticket
- Four tools — query-incidents, create-incident, query-cmdb, search-kb — cover the core ITSM workflow
- CMDB integration lets agents check asset dependencies before recommending changes or fixes
The 2.8-Hour Manual Ticket Tax
Every incident costs IT teams 2.8 hours of manual navigation: opening ServiceNow, searching for related tickets, checking CMDB dependencies, reviewing change history, and updating status fields. For a team handling 50 incidents per day, that's 140 hours of repetitive ticket work.
This FastMCP server exposes ServiceNow's ITSM APIs as MCP tools: agents create, update, and query incidents; manage change requests with approval workflows; query the CMDB for asset dependencies; and search the knowledge base for resolution patterns.
File 1: src/servicenow-mcp.ts — ServiceNow MCP Server
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import { z } from "zod";
import httpx from "httpx";
const instance = process.env.SERVICENOW_INSTANCE || "https://your-instance.service-now.com";
const auth = {
username: process.env.SERVICENOW_USERNAME || "",
password: process.env.SERVICENOW_PASSWORD || ""
};
const server = new McpServer({
name: "servicenow-itsm",
version: "1.0.0"
});
async function snGet(table: string, query: string = "", limit: number = 20) {
const params = new URLSearchParams({ sysparm_query: query, sysparm_limit: String(limit) });
const resp = await fetch(`${instance}/api/now/table/${table}?${params}`, {
headers: { "Accept": "application/json" },
// Basic auth via base64
});
return resp.json();
}
async function snPost(table: string, data: object) {
const resp = await fetch(`${instance}/api/now/table/${table}`, {
method: "POST",
headers: { "Content-Type": "application/json", "Accept": "application/json" },
body: JSON.stringify(data)
});
return resp.json();
}
async function snPatch(table: string, sysId: string, data: object) {
const resp = await fetch(`${instance}/api/now/table/${table}/${sysId}`, {
method: "PATCH",
headers: { "Content-Type": "application/json", "Accept": "application/json" },
body: JSON.stringify(data)
});
return resp.json();
}
// --- Tool 1: Query Incidents ---
server.tool(
"query-incidents",
"Query ServiceNow incidents by priority, state, or category",
{
priority: z.enum(["1", "2", "3", "4"]).optional().describe("Priority filter: 1=Critical, 2=High, 3=Moderate, 4=Low"),
state: z.enum(["1", "2", "3", "6", "7"]).optional().describe("State: 1=New, 2=In Progress, 3=On Hold, 6=Resolved, 7=Closed"),
category: z.string().optional().describe("Category filter, e.g. Hardware, Software, Network"),
limit: z.number().default(10).describe("Max results")
},
async ({ priority, state, category, limit }) => {
const filters: string[] = [];
if (priority) filters.push(`priority=${priority}`);
if (state) filters.push(`state=${state}`);
if (category) filters.push(`category=${category}`);
const query = filters.join("^");
const result = await snGet("incident", query, limit);
const incidents = (result.result || []).map((r: any) => ({
number: r.number,
short_description: r.short_description?.substring(0, 150),
state: r.state,
priority: r.priority,
assigned_to: r.assigned_to?.display_value || "Unassigned",
sys_id: r.sys_id
}));
return { content: [{ type: "text", text: JSON.stringify({ count: incidents.length, incidents }, null, 2) }] };
}
);
// --- Tool 2: Create Incident ---
server.tool(
"create-incident",
"Create a new ServiceNow incident with structured fields",
{
short_description: z.string().describe("Brief summary of the incident"),
description: z.string().describe("Detailed description"),
priority: z.enum(["1", "2", "3", "4"]).default("3"),
category: z.string().default("Software"),
assignment_group: z.string().optional().describe("Assignment group name")
},
async ({ short_description, description, priority, category, assignment_group }) => {
const result = await snPost("incident", {
short_description,
description,
priority,
category,
assignment_group,
state: "1" // New
});
const inc = result.result;
return {
content: [{ type: "text", text: JSON.stringify({
number: inc.number,
sys_id: inc.sys_id,
message: `Incident ${inc.number} created successfully`
}, null, 2) }]
};
}
);
// --- Tool 3: Query CMDB ---
server.tool(
"query-cmdb",
"Query the CMDB Configuration Management Database for assets",
{
table: z.string().describe("CMDB table: cmdb_ci_server, cmdb_ci_database, cmdb_ci_network"),
query: z.string().optional().describe("Filter query, e.g. name=prod-db-01"),
limit: z.number().default(10)
},
async ({ table, query, limit }) => {
const result = await snGet(table, query || "", limit);
const assets = (result.result || []).map((r: any) => ({
name: r.name,
ip_address: r.ip_address || "N/A",
os: r.os || "N/A",
status: r.operational_status,
sys_id: r.sys_id
}));
return { content: [{ type: "text", text: JSON.stringify({ count: assets.length, assets }, null, 2) }] };
}
);
// --- Tool 4: Search Knowledge Base ---
server.tool(
"search-kb",
"Search ServiceNow Knowledge Base articles",
{
query: z.string().describe("Search query for KB articles"),
limit: z.number().default(5)
},
async ({ query, limit }) => {
const result = await snGet("kb_knowledge", `LIKE${query}`, limit);
const articles = (result.result || []).map((r: any) => ({
number: r.number,
title: r.title,
short_description: r.short_description?.substring(0, 200),
kb_knowledge_base: r.kb_knowledge_base?.display_value,
sys_id: r.sys_id
}));
return { content: [{ type: "text", text: JSON.stringify({ count: articles.length, articles }, null, 2) }] };
}
);
// --- Start ---
const transport = new StdioServerTransport();
await server.connect(transport);
console.error("ServiceNow ITSM MCP Server running on stdio");
File 2: .env.example
SERVICENOW_INSTANCE=https://your-instance.service-now.com
SERVICENOW_USERNAME=admin
SERVICENOW_PASSWORD=your-password
Claude Desktop Configuration
{
"mcpServers": {
"servicenow": {
"command": "npx",
"args": ["tsx", "src/servicenow-mcp.ts"],
"env": {
"SERVICENOW_INSTANCE": "https://your-instance.service-now.com",
"SERVICENOW_USERNAME": "admin",
"SERVICENOW_PASSWORD": "your-password"
}
}
}
}
Production Reality Check
- Authentication: Use OAuth 2.0 with ServiceNow's token-based auth; avoid basic auth in production
- Table ACLs: Agent inherits the API user's access; use a dedicated integration user with restricted roles
- Rate limits: ServiceNow allows ~100 API calls/minute; implement request queuing
- Pagination: Use sysparm_offset for large result sets beyond the limit
- Audit logging: All API calls are logged in ServiceNow's sys_audit table for compliance
Setup Commands
# Install dependencies
npm init -y
npm install @modelcontextprotocol/sdk zod
npm install -D tsx typescript @types/node
# Run the server
npx tsx src/servicenow-mcp.ts
By Deepak Bagada, CEO at SaaSNext & Principal AI Architect.
Find more enterprise MCP integrations in our MCP Directory and explore Sentry error triage and Jira hybrid MCP servers.
Last tested: August 2026 with Node v22, TypeScript 5.5, and latest SDK releases.
Enjoyed this breakdown? Get our morning dispatch in your inbox.
Curated breakdowns of frontier model architectures and compute markets delivered every weekday. Zero fluff.
Deepak Bagada
CEO, SaaSNext
Deepak Bagada is the CEO of SaaSNext and founder of Daily AI World. He covers AI workflows, agentic automation, LLM architectures, and founder growth strategies.
RAG in 2026: When Vector Search Hits the Wall and What Comes Next
Next Story →Build a CloudWatch Observability MCP Server for Agentic Infrastructure Monitoring in 2026
Related Intelligence Analysis
Vercel AI SDK Tool Calling React: 5 Steps (2026)
Vercel AI SDK tool calling React integration is a programming pattern that executes server-side functions based on large language model decisions and streams the results to a React frontend. By combining streamText with...
Fact-Density vs. Word Count: The New SEO for 2026
Fact Density is the ratio of verifiable, unique information to the total word count of a piece of content. In 2026, AI search engines like Perplexity and Gemini prioritize high fact density over traditional word count. A...
NVIDIA Audex vs Qwen3.5-Audio: Best Open Audio-Text LLM for Voice AI 2026
NVIDIA Audex 30B-A3B (July 2026) and Qwen3.5-35B-A3B are the two leading open audio-text LLMs. Audex uniquely handles both audio understanding and generation in a single model while preserving text intelligence. Qwen3.5-...