AI 에이전트가 Table API를 통해 ServiceNow 인시던트를 생성, 검색 및 업데이트할 수 있게 해주는 작은 MCP 서버입니다. 공식 Python mcp 패키지(FastMCP 스타일 데코레이터), httpx 및 python-dotenv로 구축되었습니다. stdio를 통해 실행됩니다.
모든 것은 단일 파일 server.py에 있습니다.
Python 3.12+ 및 uv가 필요합니다.
uv sync인스턴스에 대한 기본 인증 자격 증명을 사용하여 server.py 옆에 .env를 만듭니다(이는 git에서 무시됩니다). Personal Developer Instance가 잘 작동합니다:
SNOW_INSTANCE_URL=https://devXXXXXX.service-now.com
SNOW_USERNAME=admin
SNOW_PASSWORD=your-password
# optional, default 15
SNOW_TIMEOUT_SECONDS=15서버를 실행합니다(서버는 stdin/stdout을 통해 MCP를 사용하므로 표시할 내용이 없습니다. MCP 클라이언트가 실행해야 합니다):
uv run server.py.venv와 .env가 선택되도록 프로젝트 디렉토리와 함께 uv를 가리키도록 클라이언트를 설정합니다:
{
"mcpServers": {
"servicenow": {
"command": "uv",
"args": ["--directory", "/absolute/path/to/servicenow-mcp", "run", "server.py"]
}
}
}Claude Code의 경우: claude mcp add servicenow -- uv --directory /absolute/path/to/servicenow-mcp run server.py.
클라이언트 환경에 이미 설정된 환경 변수는 .env보다 우선합니다.
Related MCP server: sn-mcp-bridge
도구 | 매개변수 | 기능 | 반환값 |
|
| 새 인시던트를 생성합니다. |
|
|
| 인시던트를 최신순으로 나열합니다. |
|
|
| 번호로 인시던트를 조회하고, 작업 메모를 추가하거나 상태를 설정합니다. |
|
모든 도구는 스택 추적 대신 읽을 수 있는 도구 오류(에이전트에 isError로 표시됨)를 발생시킵니다. 예:
인증 실패 (401): SNOW_USERNAME 및 SNOW_PASSWORD를 확인하세요. ServiceNow 응답: 사용자가 인증되지 않았습니다 …
https://…에 연결할 수 없습니다: [Errno 8] nodename nor servname provided … SNOW_INSTANCE_URL 및 네트워크 연결을 확인하세요.
https://…와 통신하는 동안 15초 후 시간 초과되었습니다. 인스턴스가 최대 절전 모드(PDI)이거나 연결할 수 없습니다.
인시던트 'INC9999999'를 https://…에서 찾을 수 없습니다.
금지됨 (403): 사용자에게 권한이 없거나 … 비즈니스 규칙에 의해 업데이트가 거부되었습니다. ServiceNow 응답: …
(예: 해결 필드 없이 인시던트를 해결하는 경우).
ServiceNowClient 는 단일 httpx.AsyncClient(기본 인증, JSON 헤더, 시간 초과)를 래핑하고 create_record / query_records / get_record_by_number / update_record를 노출합니다. 모든 오류는 사람이 읽을 수 있는 메시지와 함께 ServiceNowError로 변환됩니다.
도구는 get_client()를 통해 클라이언트를 가져옵니다. 테스트에서 set_client(fake)를 호출하여 네트워크에 접촉하지 않고 모의 객체로 교체합니다.
레코드는 sysparm_display_value=all로 가져오므로 상태/긴급 레이블은 하드코딩된 맵 대신 인스턴스 자체에서 가져옵니다(친숙한 이름 별칭은 입력 전용입니다).
mcp 2.x는 FastMCP를 MCPServer로 이름을 변경했습니다. server.py는 둘 중 존재하는 것을 가져오므로 1.x와 2.x 모두에서 실행됩니다.
라이브 인스턴스에 대해 실행합니다(하나의 인시던트를 작성합니다):
uv run python -c "
import asyncio, server
async def main():
c = await server.create_incident('MCP server smoke test', 'created by smoke test')
print(c)
print(await server.query_incidents(query_text='MCP server smoke test'))
print(await server.update_incident(c['number'], work_notes='hello from MCP'))
asyncio.run(main())
"Create a new ServiceNow incident and return its number and sys_id.
Use this when a user reports a problem that should be tracked as a ticket.
Args:
short_description: One-line summary of the issue (required, shown in lists).
description: Longer free-text details, steps to reproduce, impact, etc.
urgency: "1" (High), "2" (Medium) or "3" (Low). Defaults to "3".
Returns:
{"number": "INC0010001", "sys_id": "...", "short_description": ..., "state": ..., "urgency": ...}
The `number` is what humans reference; `sys_id` is the stable record id.
| Name | Required | Description | Default |
|---|---|---|---|
| urgency | No | 3 | |
| description | No | ||
| short_description | Yes |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations provided, so description carries full burden. It discloses that the tool creates a record and returns specific fields. It could add potential side effects or permissions needed, but is adequate.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Well-structured, front-loaded with clear purpose and usage. Each sentence adds value, no fluff. Parameters and return value explained efficiently.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Tool has 3 parameters and output schema exists. Description covers parameter semantics, return format, and usage context. No gaps given complexity.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 0%, so description must compensate. It does so by explaining each parameter's purpose and format (e.g., urgency values, description use). Adds meaning beyond schema (e.g., 'shown in lists'). Could note default for short_description.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool creates a ServiceNow incident and returns its number and sys_id. It distinguishes itself from siblings by specifying creation, not querying or updating.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly states when to use this tool: 'when a user reports a problem that should be tracked as a ticket.' No exclusion of siblings but clear context for usage, and siblings are named for differentiation.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List ServiceNow incidents, newest first, optionally filtered by text and/or state.
Use this to find existing tickets before creating a new one, to check on a
ticket by number, or to answer "what incidents are open about X?".
Args:
query_text: What to search for. An incident number ("INC0010001") matches
exactly; anything else runs ServiceNow's keyword search over the
incident's text fields, falling back to a substring match on
short_description/description. Leave empty to list the most recent incidents.
state: Filter by state. Accepts a numeric code ("1".."8") or a name such as
"New", "In Progress", "On Hold", "Resolved", "Closed", "Canceled".
Leave empty for any state.
limit: Maximum number of incidents to return (1-100). Defaults to 5.
Returns:
A list of {"number", "sys_id", "short_description", "state", "state_code",
"urgency", "created_on"}; an empty list means nothing matched.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | ||
| state | No | ||
| query_text | No |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description fully discloses behavioral details: search semantics (exact number matching vs. keyword search with substring fallback), state filter format (numeric or name), limit behavior with default, and return format. It also explains the meaning of an empty list. This is comprehensive and goes beyond a simple 'lists incidents.'
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is organized into clear sections: a one-line summary, usage guidance, parameter details, and return format. Every sentence serves a purpose, from explaining search fallback to specifying defaults. It is appropriately sized given the behavioral complexity.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The description is complete for a read-only query tool: it defines input semantics, output structure, and edge cases (empty list, empty query). The presence of an output schema is complemented by a human-readable description of the return fields. Given the sibling tools, it clearly fits into the workflow as the search/read operation. No gaps are apparent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, so the description must fully explain parameters. The 'Args' section does exactly that: query_text, state, and limit are each described with accepted formats, defaults, and special behavior (e.g., prefix matching for incident numbers, state code ranges). This adds significant meaning beyond the raw schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states 'List ServiceNow incidents, newest first, optionally filtered by text and/or state.' This uses a specific verb, names the resource (ServiceNow incidents), and distinguishes it from sibling tools like create_incident and update_incident by focusing on listing/querying.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly provides use cases: 'Use this to find existing tickets before creating a new one, to check on a ticket by number, or to answer "what incidents are open about X?".' This gives clear when-to-use guidance. It implies when-not-to-use (e.g., before creating) but does not explicitly name alternatives or exclude other tools, so it stops short of a 5.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Update an existing incident by number: append a work note and/or change its state.
Use this to record progress on a ticket or move it through its lifecycle.
At least one of `work_notes` or `state` must be provided.
Args:
number: The incident number, e.g. "INC0010001".
work_notes: Text to append to the (internal) work notes journal.
state: New state as a numeric code ("1".."8") or a name such as "In Progress",
"On Hold", "Resolved", "Closed". Note: resolving/closing on most instances
also requires resolution fields and will be rejected by ServiceNow if missing.
Returns:
{"number", "sys_id", "changed": {field: new value, ...}, "state": <current state>}
describing exactly what was applied.
| Name | Required | Description | Default |
|---|---|---|---|
| state | No | ||
| number | Yes | ||
| work_notes | No |
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description bears full responsibility for behavioral disclosure. It clearly states that this is a mutating operation (update), explains the effect of each parameter, and warns that resolving/closing may be rejected by the backend. It does not mention idempotency, error handling for missing incidents, or authentication requirements, but the provided details are sufficient for safe invocation.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is concise and well-structured. It starts with a one-sentence summary of the action, then a one-sentence usage context, then a constraint, then a clear Args/Returns section. Every sentence adds unique information; there is no redundancy.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
The description covers all parameters, provides a return format, and includes a key behavioral caveat (rejection on missing resolution fields). It is almost complete for a mutation tool with no annotations, but it does not specify what happens if the incident number does not exist or whether the operation is idempotent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 0%, so the description must fully explain each parameter. It does so: number is given an example ('INC0010001'), work_notes is described as 'append to the (internal) work notes journal', and state is documented with both numeric codes and names, plus a note about additional fields required by ServiceNow. This adds significant semantic value beyond the bare schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description specifies a clear verb ('update'), resource ('existing incident'), and method ('by number'). It explicitly lists the two possible actions (append work notes and/or change state), which distinguishes it from siblings create_incident and query_incidents.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description advises when to use the tool ('record progress on a ticket or move it through its lifecycle') and states a critical constraint ('at least one of work_notes or state must be provided'). It also warns about potential rejection when resolving/closing if resolution fields are missing. However, it does not explicitly contrast with alternatives like create_incident or query_incidents.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.0create_incidentquery_incidentsupdate_incidentScored across 3 tools
Each tool targets a distinct operation: create, query (list/search), and update. There is no overlap in purpose; an agent can clearly differentiate when to use each.
All tools follow a consistent verb_noun pattern using snake_case (create_incident, query_incidents, update_incident), with plural nouns for collection queries and singular for single-item operations, which is predictable.
With 3 tools, the set is minimal but covers the core incident lifecycle (create, read/search, update). It is slightly thin—missing a dedicated get_incident or delete_incident—but still well-scoped for a focused incident-only server.
The tools cover create, list/search, and update, but lack an explicit 'get by ID' tool (though query_incidents can filter by number) and a delete/close operation. The update tool warns about missing resolution fields for closing, indicating gaps in full lifecycle support.
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
MCP server for OpenAI API (chat completions, image generation, embeddings) via AceDataCloud
MCP server for AI dialogue using various LLM models via AceDataCloud
MCP server that lets AI assistants use all OneSchema features exposed via the public API.
