A desk lit by a warm lamp in the evening, a laptop open to a support-ticket queue with a coffee mug beside it

ChatGPT Work + ServiceNow: What the Integration Can (and Can't) Do in 2026

Partly — there’s no official OpenAI connector for ServiceNow, but you can wire ChatGPT Work to ServiceNow’s own Now Assist MCP Server as a custom MCP connector. ChatGPT Work ships built-in connectors for Google Drive, Gmail, Google Calendar, Outlook, SharePoint, OneDrive, Teams, Dropbox, Box, GitHub, HubSpot, Linear, Canva, and Notion — and OpenAI’s own docs say they deliberately skipped services that already publish a remote MCP server, which is exactly why ServiceNow isn’t in that list. ServiceNow ships its own, and you point ChatGPT at it. The catch: it’s not a paste-a-URL freebie. The Now Assist MCP Server went GA in the Zurich Patch 4 release, it exposes Now Assist skills (not raw tables) as tools, it authenticates over OAuth, and it requires a Now Assist SKU — and every skill call is metered on ServiceNow’s side while the agent run is metered on ChatGPT’s side. And like every ChatGPT connection, it only moves inside a session you’re driving — between chats, nothing is watching your ticket queue.

Here’s what the ChatGPT Work + ServiceNow integration actually does, how to stand it up, where the ceiling is, and what to use when you want ITSM work that runs without you in the chat.

What ChatGPT Work can actually do with ServiceNow

Once ServiceNow’s Now Assist MCP Server is registered as a custom connector, ChatGPT Work reaches ServiceNow through Now Assist skills exposed as tools — not arbitrary read/write on every table. What you get depends on which skills your admin publishes:

  • Query incidents and records conversationally. “Summarize the P1 incidents opened in the last 24 hours and who owns them” — answered from live instance data, no report builder, as long as a skill exposes that read.
  • Take governed write actions. Because the skills run on the authorizing user’s behalf, an agent can do things like draft an incident summary, update a record, or trigger a packaged flow — scoped to what that user’s role and the published skill permit.
  • Work custom-built skills. Anything your team packages with the Now Assist Skill Kit (NASK) — a knowledge-article lookup, a change-request helper, a CMDB query — shows up as a callable tool. Individual scripts or flows have to be wrapped as a skill first; the MCP server doesn’t expose raw APIs.
  • Cross the rest of your stack in one run. With ChatGPT Work’s agent runs, a session can pull an incident from ServiceNow, cross-reference a GitHub issue, and draft the Outlook update in one long, autonomous pass — a triage sweep before your morning standup. Still a run you start.

How to set it up

This is an admin job on the ServiceNow side, not a one-click add:

  1. On the ServiceNow instance: be on a release that includes the Now Assist MCP Server — it’s GA in Zurich Patch 4 (Yokohama Patch 11 is the alternate) — and hold a qualifying Now Assist SKU (Pro Plus or Enterprise Plus), per ServiceNow’s MCP enablement guide.
  2. Publish the skills you want exposed. Decide which Now Assist skills — and any custom NASK skills — become tools. Raw scripts and flows have to be packaged as skills to be reachable.
  3. Configure OAuth on the MCP Server endpoint so the external client authenticates as the calling user and inherits that user’s role permissions.
  4. In ChatGPT Work, add a custom MCP connector and point it at the ServiceNow MCP Server endpoint — OpenAI’s MCP connector docs cover the server_url flow, plus the Secure MCP Tunnel if your instance sits behind a firewall.
  5. Test with something read-only: “list the incidents assigned to my group that are still open.”

The limits that matter

  • No official OpenAI connector — you’re on the custom-MCP path. OpenAI explicitly deprioritized services that already ship a remote MCP server, so ServiceNow lives outside the curated connector gallery. That means an admin-configured MCP endpoint, not a name you pick from a list.
  • It doesn’t run on triggers. There’s no “when a P1 incident is opened, page the on-call and draft the comms” or “when a change request is approved, kick off the tasks.” ChatGPT Work touches ServiceNow when you prompt it — an incident can breach SLA overnight and nothing moves.
  • Session-bound, even in agent mode. ChatGPT Work’s agent runs are long and autonomous, but they’re manually started and metered against your plan’s allowance. Nothing runs between chats; close the session and the watch ends.
  • Metered on both sides. On top of ChatGPT’s run metering, ServiceNow meters the skills: a Now Assist skill invoked over MCP consumes its standard assist count plus one additional assist per call. A chatty agent burns assists.
  • You only reach what’s been packaged as a skill. The MCP server exposes Now Assist and custom NASK skills, not the full Table API. If a capability isn’t wrapped as a published skill, ChatGPT can’t call it.
  • The follow-through stops at the chat. ChatGPT can draft the incident summary or the change recap; it won’t then send it from your mailbox, book the review call, and post the update to your ops channel as one unbroken motion.

If you want ServiceNow-adjacent work that runs on its own: Carly

ITSM is trigger-shaped. Incidents open at 3am, change requests get approved mid-afternoon, SLAs tick down over the weekend. The moment you want the response to happen on the event — a new P1 summarized and the owner emailed within minutes, a nudge when a ticket goes stale, a Monday queue digest built and sent — you’ve crossed past what a chat session, even a metered agent run, is for.

That’s where Carly fits. Carly is an AI executive assistant built to act on triggers, not just answer when you ask:

  • Fires on events and schedules, 24/7, in the cloud. When a ticket is opened or a change is approved, Carly reacts — summarizes it, emails the owner, updates the record, posts to Slack — while your laptop is closed.
  • Actually sends, not just drafts. Carly drafts and sends email across Gmail and Outlook, books meetings, updates tickets, tasks, and CRM records, and records meetings — the follow-through that stops at the chat with ChatGPT.
  • Builds the workflow by interviewing you. Tell Carly “when a P1 incident opens, summarize it, email the on-call lead, and post to #ops” in plain English; it interviews you and builds the workflow — no admin center, no prompt engineering.
  • Connects to your ServiceNow instance with your own key. ServiceNow isn’t a native Carly integration yet, so you connect it BYO-key — paste your API key on carlyassistant.com/integrations — alongside 200+ native tools across 40+ categories. See integrations.

AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited.

ChatGPT Work vs Carly

ChatGPT Work (ServiceNow MCP)Carly
Query incidents and recordsYes, in a sessionYes
Take write actions in ServiceNowYes, via published skillsYes (BYO-key)
Official one-click connectorNo (custom MCP, admin-configured)BYO-key setup
Reacts to a new incident by itselfNoYes, on the trigger
Monday queue digest, on scheduleNoYes
Emails the owner / on-call as part of the flowNo (drafts only)Yes (Gmail + Outlook)
Runs without a session openNo (agent runs started + metered)Yes (cloud, 24/7)
MeteringChatGPT run + ServiceNow assists (+1 per skill)AI agents from $35/mo; non-AI steps free
Requires a Now Assist SKUYes (Pro Plus / Enterprise Plus)No
SetupAdmin publishes skills, OAuth, custom MCP URLDescribe it in plain English

ChatGPT Work’s ServiceNow connection is an ITSM copilot you steer in a chat. Carly is a teammate that works the queue while you’re in a change review.

Frequently Asked Questions

Does ChatGPT Work integrate with ServiceNow?

Yes, but not through an official OpenAI connector. ChatGPT Work’s built-in connectors cover Google Drive, Gmail, Google Calendar, Outlook, SharePoint, OneDrive, Teams, Dropbox, Box, GitHub, HubSpot, Linear, Canva, and Notion — ServiceNow isn’t among them. Instead you connect ServiceNow’s own Now Assist MCP Server as a custom MCP connector. OpenAI says it deliberately skipped services that already publish a remote MCP server, which is why ServiceNow ships its own instead.

Is there an official ServiceNow MCP server for ChatGPT?

ServiceNow ships a native Now Assist MCP Server, GA in the Zurich Patch 4 release, that exposes Now Assist and custom NASK skills as tools to any external MCP client — Claude, ChatGPT, Cursor, and others. It authenticates over OAuth as the calling user and requires a Now Assist SKU (Pro Plus or Enterprise Plus). It’s official from ServiceNow, not from OpenAI.

Can ChatGPT Work update ServiceNow incidents automatically when they’re opened?

No. ChatGPT Work only touches ServiceNow when you prompt it, and even its agent runs are manually started and metered — nothing fires on a ServiceNow event. For “when a P1 opens, summarize it and email the on-call lead,” you need a trigger-based assistant like Carly, which runs in the cloud 24/7 and connects to ServiceNow with your own API key.

How much does the ServiceNow side cost to use over MCP?

Beyond ChatGPT’s own run metering, ServiceNow requires a Now Assist SKU and meters usage: a Now Assist skill invoked over the MCP server consumes its standard assist count plus one additional assist per call, per ServiceNow’s MCP enablement guidance. So a talkative agent can run up assist consumption quickly.


More: ChatGPT MCP · Can ChatGPT send emails · ChatGPT personal assistant · Best AI agents for productivity

Ready to automate your busywork?

Carly schedules, researches, and briefs you—so you can focus on what matters.

See what people say

"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.

Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.

On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."

Gus Ibrahim, Founder & Director, IHR