Using MCP Server for the Windchill AI Assistant
The Windchill AI Assistant plugin includes a Model Context Protocol (MCP) server. This server enables external AI clients, such as Microsoft 365 Copilot, Creo+, or custom enterprise AI solutions, to invoke Windchill AI Assistant capabilities through a standardized protocol. By connecting an MCP client to the plugin, you can access these capabilities directly from your existing tools.
MCP Server Endpoint
An MCP client connects to the following endpoint to discover and call the available tools:
https://<windchill-host>/Windchill/WCPlugins/wcaiassistant/mcp
Where, <windchill-host> is the fully qualified host name of your Windchill server.
Typical Workflow
The following workflow describes how an MCP client interacts with the plugin:
1. The MCP client authenticates against the Windchill Apache server by using a supported authentication method.
2. The client connects to the MCP endpoint and discovers the list of available tools.
3. The client calls a tool with the required parameters.
4. The plugin runs the tool as the authenticated user, enforces Windchill access rights, and returns the response to the client.
MCP Tools
The MCP server exposes the following tools. The tools that a client can discover depend on the plugin configuration and on the calling user's license.
|
Tool
|
Description
|
|---|---|
|
windchill_assist_help
|
Provides answers about Windchill features, configuration, usage, and troubleshooting. Uses content from the Windchill Help Center and PTC Technical Support only.
|
|
windchill_assist_documents
|
Answers questions that may require both document interpretation and Windchill platform guidance. Combines information from the Windchill Document Vault, the Help Center, and Technical Support articles.
|
|
windchill_assist
|
Answers open‑ended questions that require reasoning, synthesis, or decision‑making. Combines information from the Windchill Document Vault, the Help Center, and Technical Support articles.
|
|
windchill_search_documents
|
Returns relevant text chunks and associated metadata from the Windchill Document Vault. No AI reasoning is performed.
|
|
windchill_search_help
|
Returns relevant article chunks and links from the Windchill Help Center and PTC Technical Support. No AI reasoning is performed.
|
|
windchill_summarize_document
|
Generates a concise, factual summary of a specific Windchill document. Requires the document's Object ID (OID).
|
 |
The list of tools is determined when the client connects. If the Use PTC Help and Support Index opt‑in is enabled or disabled, the MCP client must reconnect to refresh the tool list.
|
Authentication
The MCP server uses the same authentication mechanisms as other Windchill REST APIs:
• Basic authentication.
• Single Sign‑On (SSO) for both on‑premises and Windchill+ deployments.
Cookie‑based authentication is not supported, because the MCP protocol does not support it.
Each tool call runs as the authenticated user. Windchill enforces access rights, so users can retrieve only the data they are authorized to view.
Connecting an MCP Client to the MCP Server
The following example shows how to connect Visual Studio Code by using basic authentication. The same parameters apply to any MCP‑compliant client. Consult your client's documentation for the equivalent configuration steps.
To connect Visual Studio Code to the MCP server by using basic authentication, complete the following steps.
1. In Visual Studio Code, open the workspace folder that you want to use as the MCP project.
2. In the workspace, create a .vscode folder if it does not already exist.
3. In the .vscode folder, create a file named mcp.json.
4. Add the following configuration to mcp.json. Replace <windchill-host> with your Windchill host name and <base64-encoded-credentials> with your encoded credentials.
{
"servers": {
"wnc-ai-assistant-mcp": {
"url": "https://<windchill-host>/Windchill/WCPlugins/wcaiassistant/mcp",
"type": "http",
"headers": {
"Authorization": "Basic <base64-encoded-credentials>",
"X-XSRF-TOKEN": "<xsrf-token>",
"Cookie": "XSRF-TOKEN=<xsrf-token>"
}
}
}
}
"servers": {
"wnc-ai-assistant-mcp": {
"url": "https://<windchill-host>/Windchill/WCPlugins/wcaiassistant/mcp",
"type": "http",
"headers": {
"Authorization": "Basic <base64-encoded-credentials>",
"X-XSRF-TOKEN": "<xsrf-token>",
"Cookie": "XSRF-TOKEN=<xsrf-token>"
}
}
}
}
|
|
The X-XSRF-TOKEN header value and the XSRF-TOKEN cookie value can be any valid token. However, both values must be identical.
|
5. Save the file.
6. In your MCP client, connect to the wnc-ai-assistant-mcp server and verify that the list of available tools is displayed.
To use SSO, configure the MCP client to present the SSO token issued by your identity provider instead of using the basic authentication header.