Nintex Process Manager MCP Server is now in open beta, which means for the first time, you can point an AI assistant (e.g. Claude or Microsoft Copilot) directly at your live Process Manager instance and ask it questions in plain language. Rather than needing to search repositories, navigate systems, or rely on subject matter experts, Process Manager MCP Server makes approved process knowledge available directly within your flow of work via your preferred AI assistant.
This open beta focuses on secure, read-only discovery and retrieval, enabling your organization to unlock greater value from their existing process investments while maintaining governance and control. But you cannot prove improvement without a baseline.
So before you use it to build your onboarding assistant, a self-service help agent, or an automation shortlist, we recommend you use it to do something less exciting but more valuable: Find out what state your process library is actually in.
Why start with a health check?
Many conversations about AI agents jump straight to fun automation: “Let’s build an assistant that answers volunteer questions,” or “let’s connect this to a workflow.” Starting that way can potentially bring you to the right destination, but it is not the correct first move as it brings in irrelevant information without auditing the content.
An agent built on top of a process library with gaps, orphaned owners, and five-year-old content doesn’t fix those problems: It provides irrelevant and out-of-date content that may mislead the user. It’ll answer confidently from whatever’s there; gaps included.
So, the first job for a Process Champion is the same as it’s always been: Knowing the state of your library before you point anything powerful at it.
This exercise enables you to capture a baseline to show the positive impact you’ve made on the team’s efficiency. Once you’ve established a baseline, you can showcase in a tangible way that you have increased process adoption and utilization, while helping reduce time spent searching for process information.
What you’ll need before you start
| Requirement | Detail |
| Admin enablement | A Process Manager admin must turn on MCP under Configure > Admin > Enable Process Manager MCP |
| Your MCP URL | Format: https://<region>-pmp.promapp.com/mcp/<tenant name> |
| Region code | au → prd-aue · us → prd-wu3 · eu → prd-neu · ca → prd-cac · demo → dmo-wu3 |
| Login | Your normal Process Manager username/password, or SSO if enabled |
Example: if your Process Manager site is https://au.promapp.com/mycompany, your MCP URL is https://prd-aue-pmp.promapp.com/mcp/mycompany.
One thing worth knowing before you connect anything: The agent only sees what the authenticated user can see. Process and search results are filtered by the permissions of whoever logs in. If you want a full library view for a health check, connect using an account with broad read access and not a restricted departmental login.
The available tools let an agent search and retrieve processes and documents, pull full process details, browse the group hierarchy, and list processes. There’s no write access just yet (coming soon) and no visual process map rendering yet either. Set that expectation early so nobody’s disappointed when the agent describes a process instead of drawing it.
Connecting Claude
If you’re on Team or Enterprise licenses: An Owner adds the connector once, centrally, at Organization Settings > Connectors > Add > Custom > Web, pastes in the MCP URL, and everyone else just connects to it individually from Customize > Connectors.
If you’re on Pro or Max licenses: Go to Customize > Connectors > Add custom connector, paste in your MCP URL, and authenticate.
Either way, once it’s added, you switch it on per conversation from the “+” menu under Connectors. It won’t run in the background across every chat by default: You must turn it on when you want the agent working against Process Manager.
Connecting Copilot Studio
In your agent’s Tools page, select Add a tool > New tool > Model Context Protocol. Enter the server name, a clear description (the orchestrator uses this to decide when to call it), and your MCP URL. Whether your tenant uses SSO or credentials for authentication, you’ll configure OAuth 2.0 authentication here rather than “None.”
The single most important instruction
Whichever tool you use, give your agent this instruction before anything else: Tell it to treat Process Manager as the source of truth for process questions, and not to supplement or override that with general web knowledge.
That will look something like:
“For any question about our processes, policies, or organizational structure, use information retrieved from Process Manager. Do not supplement with general knowledge or web search and if Process Manager doesn’t have an answer, say so directly.”
This single instruction is what stops the agent from confidently inventing a process that doesn’t exist or blending your actual approved procedure with a generic best-practice answer it picked up from other external sources. It’s the difference between an assistant your organization can trust and one that sounds right but isn’t.
Health check prompt library
Next, run these prompts directly in a Claude or Copilot conversation with the Process Manager connector switched on. They’re grouped by what they surface. Adjust the wording to your own terminology (directorate, business unit, or department).
1. Coverage & structure
- “List all process groups in our group hierarchy, and how many processes sit under each one.”
- “Which groups have fewer than 3 documented processes? These are candidates for either genuine simplicity or under-documentation, flag both possibilities.”
- “Search for processes related to [core function, e.g. ‘onboarding’, ‘reimbursement’, ‘incident reporting’]. How many versions or near-duplicates come back?”
2. Ownership & accountability
- “Pull the full detail for [process name] and tell me who the listed owner is.”
- “Search for processes with no clear owner or approver noted in their detail, list what you find.”
- “For [group name], list every process and its current owner, so I can spot-check for people who’ve since changed roles.”
3. Freshness & risk
- “Search documents and policies related to [regulatory area]. Which ones look like they haven’t been reviewed recently, based on content or references to outdated systems/roles?”
- “Pull [process name] in full, does the content reference any system, form, or role name we no longer use?”
4. Automation & AI-readiness signals
- “Review [process name] and identify which steps are manual, repetitive, and rules-based, the kind of steps worth flagging for automation.”
- “Across [group name], which processes mention manual data entry, duplicate approvals, or handoffs between more than two people or teams?”
Note on hallucination risk: Even with the “Process Manager only” instruction in place, treat every answer as a lead, not a finding. The agent is good at retrieving and summarizing what’s there, not auditing your organization. A Process Champion or BA should sight-check anything before it goes in front of a sponsor.
Turning your findings into a scorecard
Once you’ve run the categories above, you have the makings of a simple baseline scorecard, even before you touch value or ROI. This scorecard will surface things like:
- Coverage: % of functions/teams with documented process vs known gaps
- Ownership clarity: % of processes with a confirmed, current owner
- Currency: % of processes with no obvious staleness flags
- Automation readiness: count of manual/repetitive steps identified as candidates
Four numbers, one page, taken before you do anything else. That’s your “before.”
Cadence
To maintain a healthy baseline, run the health check prompt set quarterly or whenever a new business area is onboarded to Process Manager. The Process Champion owns running the health check, and Operations and Senior BAs are the right people to sight-check the automation and risk flags before they’re actioned.
Next in this series: How to put a defensible dollar figure on a process, and how current vs future state analysis feeds the calculation.
Unlock trusted process knowledge with AI
Connecting Claude or Copilot to Nintex Process Manager gives your teams a faster way to find approved process information, uncover gaps in documentation, and establish a baseline for continuous improvement. A health check is often the simplest way to understand the current state of your process library before investing in automation or transformation initiatives.
If you’re ready to explore Process Manager MCP Server, learn more about the open beta to see how AI-powered process discovery could work for your organization.