Process Manager MCP

Important: This feature is currently in open beta. There may be incorrect results returned. If something is not functioning correctly, please raise this with support@nintex.com

Process Manager MCP is an open-source Model Context Protocol (MCP) server that connects AI Assistants/Agents (e.g. through Claude Desktop and Microsoft Copilot Studio) to your Nintex Process Manager instance. It enables you to ask questions about your processes, documents, and organizational structure in plain language, without needing to navigate within the Process Manager user interface.

Some examples of what you can do:

• Search for processes by title, activities, tasks, objectives, notes, and keywords.

• Find documents and policies stored in your Process Manager instance.

• Retrieve the full details of a specific process.

• Analyze a process to identify potential automation opportunities.

Setup steps

1. Your Promaster A Nintex Process Manager system administrator who administers a Nintex Promapp site and has rights to view and edit all areas of a Nintex Promapp site. will first need to enable the Process Manager MCP capability for your tenant by navigating to Configure > Admin > Enable Process Manager MCP.

2. Get your MCP URL for your Process Manager tenant.

   1. This would take the following format: https://<region-identifier>-pmp.promapp.com/mcp/<tenant name>.

   2. The corresponding regions are:

      1. au -> prd-aue

      2. ca -> prd-cac

      3. eu -> prd-neu

      4. us -> prd-wu3

      5. demo -> dmo-wu3

   3. For example, if the URL for your Process Manager tenant is https://au.promapp.com/mycompany, your MCP URL would be https://prd-aue-pmp.promapp.com/mcp/mycompany.

3. Go to your AI Assistant/Agent setup (e.g. Microsoft Copilot Studio) and follow your chosen instruction to connect an MCP server to your AI Assistant/Agent.

   1. The general flow should be similar to the following:

      1. Edit your AI Assistant/Agent.

      2. Create a custom MCP connection for your agent. Here is an example of creating a custom MCP connection in Microsoft Copilot:

         

         Note: Different providers may call the custom MCP connection differently.

      3. Paste the MCP URL for your Process Manager tenant from the earlier step.

      4. You will most likely be asked to log in to authenticate at this step.

      5. Finish setting up your agent with additional instructions and tools.

   2. Examples:

      1. Microsoft Copilot Studio: Connect your agent to an existing Model Context Protocol (MCP) server.

      2. Claude Desktop: Get started with custom connectors using remote MCP.

4. When you launch your agent, you may be asked to authenticate again either via providing your username and password or via Single Sign-On (SSO) if that is enabled for your Process Manager tenant.

   Note: You may be asked to authenticate again if your use of the MCP server times out.

Step-by-step video examples

Considerations

• Instruct your AI Assistant/Agent that for any process related questions it should seek information from Process Manager rather than other sources, e.g. through the web. This should limit hallucinations where your agent pulls process information from other sources.

  See the MCP FAQ.

• For prompt ideas and suggestions see Suggested agent prompts for Process Manager MCP.

Available MCP tools

The following MCP tools are executable functions that let AI models interact with Process Manager.

Tool	What it does
get_process	Retrieve complete process details for a specific process ID.
get_process_diagram	Returns an interactive process diagram (process map) for a specified process GUID. The response includes a permalink to the minimode diagram. For more information on minimode links see Share a process. Before calling this function, retrieve the process GUID from list_processes and pass it as the input value.
list_group_hierarchy	Retrieves the immediate child process groups for a given parent group. When no parentGroupUniqueId is provided, the tool returns all top-level process groups. The response is limited to a single hierarchy level and includes: A formatted tree view for easy navigation Structured JSON for integration and automation scenarios
list_processes	Returns a paginated list of all processes with meta data.
ping	Returns a response to verify the MCP server is working.
review_process_for_automation	Analyzes a process and evaluates each activity for automation potential. The analysis identifies patterns such as data entry, data movement, system lookups, document processing, approvals, notifications, and other repetitive tasks. The response includes: Per-step automation opportunities Confidence and implementation complexity ratings Recommended automation approaches (API integration, RPA, workflow automation, or dedicated agents) Optional: recommendations for dedicated-agents with suggested architectures Before calling this function, retrieve the process GUID from list_processes and pass it as the input value.
search_all	Search across all content types: processes, documents, policies, and groups.
search_documents	Find documents within your Process Manager instance.
search_group_hierarchy	Searches for process groups by name and returns all matching groups, including their subgroups and processes. The hierarchy is expanded to a maximum depth of five levels by default. Increase maxDepth to retrieve deeper levels, or set it to null to return the complete hierarchy. The response includes: A formatted tree view for easy navigation Structured JSON for integration and automation scenarios
search_processes	Search across process titles, activities, tasks, objectives, notes, and keywords.