sn-mcp-bridge

A lightweight Model Context Protocol (MCP) server that gives AI coding assistants full development capability on ServiceNow — no installation required anywhere. It runs locally via npx and connects to ServiceNow through the Table API.

ServiceNow is a record-based development platform. Script includes, business rules, client scripts, UI actions, ACLs — every development artifact is a record in a system table. There is no separate "code layer"; the Table API is the development API. That means CRUD operations through this server aren't just for querying data — they're how you build:

• insert_record into sys_script_include → create a new script include

• update_record on sys_script → modify a business rule

• query_data on sys_script_client → read all client scripts for a table

• delete_record on sys_ui_action → remove a UI action

For tasks that go beyond CRUD — testing logic, running complex GlideRecord queries, calling script includes, or multi-step transactions — execute_script provides a full server-side JavaScript runtime.

The server runs with the permissions of whatever user account you provide credentials for — it can only read/write tables and fields that user has access to. The execute_script tool requires admin credentials since it runs background scripts via sys.scripts.do.

There are plenty of open-source ServiceNow MCP servers being shared in the community. This one exists to stay simple, pure, and easy to improve — plain JS with no build step (2 source files, native fetch, Node.js 18+), and adding a tool is one server.registerTool() call.

Tools

CRUD

Schema & Discovery

Analytics

Advanced

Related MCP server: ServiceNow MCP Server

Quick Start

Environment Variables

<INSTANCE_NAME> is the subdomain from SN_INSTANCE, uppercased with hyphens replaced by underscores (e.g. https://mydev01.service-now.com → SN_MYDEV01_USERNAME). If the prefixed vars aren't set, the bridge falls back to SN_USERNAME / SN_PASSWORD.

Warning: The examples below use plaintext credentials to get you running quickly. This means your password is stored in a file on disk and visible to the AI assistant in every API call to the LLM provider. Once you've confirmed the connection works, it is highly recommended that you follow the Securing Credentials with Secretless AI instructions to move plaintext secrets out of your config!

Claude Code

Add to .mcp.json in your project root (only available in that project) or ~/.claude/claude_code_config.json (available in all projects):

{
	"mcpServers": {
		"sn_mydev01": {
			"command": "npx",
			"args": ["-y", "sn-mcp-bridge"],
			"env": {
				"SN_INSTANCE": "https://mydev01.service-now.com",
				"SN_MYDEV01_USERNAME": "your_username",
				"SN_MYDEV01_PASSWORD": "your_password"
			}
		}
	}
}

OpenAI Codex

Add to .codex/config.toml in your project root (project-only, requires a trusted project) or ~/.codex/config.toml (available in all projects):

[mcp_servers.sn_mydev01]
command = "npx"
args = ["-y", "sn-mcp-bridge"]

[mcp_servers.sn_mydev01.env]
SN_INSTANCE = "https://mydev01.service-now.com"
SN_MYDEV01_USERNAME = "your_username"
SN_MYDEV01_PASSWORD = "your_password"

VS Code (GitHub Copilot)

Add to .vscode/mcp.json in your project:

{
	"servers": {
		"sn_mydev01": {
			"type": "stdio",
			"command": "npx",
			"args": ["-y", "sn-mcp-bridge"],
			"env": {
				"SN_INSTANCE": "https://mydev01.service-now.com",
				"SN_MYDEV01_USERNAME": "your_username",
				"SN_MYDEV01_PASSWORD": "your_password"
			}
		}
	}
}

Cursor

Add to .cursor/mcp.json in your project:

{
	"mcpServers": {
		"sn_mydev01": {
			"command": "npx",
			"args": ["-y", "sn-mcp-bridge"],
			"env": {
				"SN_INSTANCE": "https://mydev01.service-now.com",
				"SN_MYDEV01_USERNAME": "your_username",
				"SN_MYDEV01_PASSWORD": "your_password"
			}
		}
	}
}

Multiple Instances

Add a separate server entry for each instance. The config format is the same as above — just repeat the pattern with a different server name and instance-specific credentials.

Securing Credentials with Secretless AI

Secretless AI stores your credentials in a secure backend and injects them at runtime via secretless-ai run.

Setup

1. Store your credentials:

npx secretless-ai secret set SN_MYDEV01_USERNAME=your_username

# Omit the value so it prompts interactively — keeps the password out of shell history
npx secretless-ai secret set SN_MYDEV01_PASSWORD

2. Update your MCP config to use secretless-ai run as a wrapper. The --only flag tells it which secrets to inject. SN_INSTANCE is not a secret and stays in the env block:

{
	"mcpServers": {
		"sn_mydev01": {
			"command": "npx",
			"args": ["-y", "secretless-ai", "run", "--only", "SN_MYDEV01_USERNAME,SN_MYDEV01_PASSWORD", "--", "npx", "-y", "sn-mcp-bridge"],
			"env": {
				"SN_INSTANCE": "https://mydev01.service-now.com"
			}
		}
	}
}

For multiple instances, repeat the pattern — store each instance's credentials under its prefixed names and add a server entry with the corresponding --only list:

{
	"mcpServers": {
		"sn_mydev01": {
			"command": "npx",
			"args": ["-y", "secretless-ai", "run", "--only", "SN_MYDEV01_USERNAME,SN_MYDEV01_PASSWORD", "--", "npx", "-y", "sn-mcp-bridge"],
			"env": {
				"SN_INSTANCE": "https://mydev01.service-now.com"
			}
		},
		"sn_myprod01": {
			"command": "npx",
			"args": ["-y", "secretless-ai", "run", "--only", "SN_MYPROD01_USERNAME,SN_MYPROD01_PASSWORD", "--", "npx", "-y", "sn-mcp-bridge"],
			"env": {
				"SN_INSTANCE": "https://myprod01.service-now.com"
			}
		}
	}
}

The config format for other editors follows the same pattern shown in Quick Start — just replace the command/args with the secretless wrapper.

Supported Backends

Alternative: protect-mcp

If your MCP configs are in global config paths (e.g. ~/.vscode/mcp.json, ~/.cursor/mcp.json), you can use protect-mcp to automatically scan and secure them in one shot:

npx secretless-ai protect-mcp --backend keychain

You can check status or revert with mcp-status and mcp-unprotect:

npx secretless-ai mcp-status
npx secretless-ai mcp-unprotect

Limitation: protect-mcp, mcp-status, and mcp-unprotect only discover global config files. They do not find workspace-level configs like .vscode/mcp.json, .mcp.json, or .codex/config.toml inside project directories. For workspace configs, use the secret set + run approach above.

For more details on Secretless AI, see the full documentation.

Requirements

• Node.js 18+ (for native fetch)

• A ServiceNow instance with REST API access

• Basic auth credentials for the instance

License

MIT