Skip to main content

Overview

MCP Sources let you connect an MCP-compatible source, such as your Mintlify docs, so Pierview can query it before writing content. The AI uses your documentation to ground articles in accurate product details: features, terminology, and how things actually work. This means generated content references your real product instead of guessing.
Your MCP server must be publicly accessible with no authentication required. Pierview queries the endpoint without credentials, so a private or auth-gated server cannot be used.

Step 1: Open the Documents page

In Pierview, go to Brand Kit → Documents. Find the MCP Sources card and click Add Source.
MCP Sources card with an Add Source button and a connected Mintlify docs source

Step 2: Add your MCP server

In the dialog, enter:
FieldWhat to enter
NameA label for the source (e.g. Mintlify Docs)
MCP Endpoint URLThe full URL of your public MCP server (e.g. https://your-docs.mintlify.app/mcp)
Add MCP Source dialog with Name and MCP Endpoint URL fields
For Mintlify docs, the MCP endpoint is found in your Mintlify dashboard under Integrations → MCP. It’s typically your docs URL with /mcp appended.
Click Add Source to save it. The source now appears in the MCP Sources card and is used as a reference whenever content is created.

How It Works

When Pierview generates an article, it queries your connected MCP sources during the research phase. Information retrieved from your documentation is prioritized when grounding the content, so the AI describes your product accurately rather than relying on general web knowledge alone. You can connect more than one source, all enabled sources are queried for each piece of content. To remove a source, click the trash icon next to it in the MCP Sources card.

Troubleshooting

The endpoint URL must start with https://. Double-check the URL is complete and points to your MCP endpoint (often ending in /mcp).
Make sure the MCP server is public and requires no authentication, Pierview queries it without credentials. If the server is auth-gated or behind a login, queries will fail and the AI will fall back to web research.