It connects your AI client to Morningstar data and research tools. Once connected, you can look up funds and securities, retrieve datapoints, review analyst research, and analyze portfolios and holdings using natural-language requests in the client you already use.

The plugin is a single install that configures the MCP Server and adds Morningstar-authored skills. Setting up manually gives you the MCP Server only. If your client supports plugins, the plugin is the quickest route.

Skills are pre-built workflows that run on top of the MCP Server. They guide your AI client through repeatable research tasks, use live Morningstar data, produce consistently formatted output, and include guardrails against inferred data. There are five: fund-screener, fund-comparison, fund-summarizer, medalist-rating-analyzer and datapoint-finder. Skills come with the plugin.

You can connect the MCP Server from Claude, ChatGPT, Microsoft Copilot Studio, the Copilot app, Microsoft Foundry and Perplexity, or from your own code. The plugin is available through the ChatGPT marketplace, and for Claude it is installed manually because it isn't yet listed in the Claude plugin marketplace.

ETFs, open-end funds, closed-end funds, collective investment trusts and stocks, across global securities. Analyst research covers stocks, open-end funds and ETFs. Fund research covers more than 20 domiciles, including the US, Canada, Japan, Australia, France, the UK and Luxembourg.

Yes. Securities and data points are available only if they're part of your Morningstar licence. Your Morningstar account manager can confirm what's included.

Self-serve trial access is available. See the Trial Access page for details.

In an AI client such as Claude, ChatGPT or Copilot, you sign in with your Morningstar credentials using OAuth 2.0 the first time you connect. From your own code, send a bearer token obtained from the Authentication API in the Authorization header. Microsoft Foundry supports both methods.

https://mcp.morningstar.com/mcp. Every platform connects to the same server.

An administrator adds Morningstar once as a custom connector, using the server URL, which makes it available to everyone in the account. Each user must still authenticate with their own Morningstar credentials.

Confirm the connector or tool was added with the server URL https://mcp.morningstar.com/mcp, then try connecting again.

Confirm your Morningstar account is licensed for the data you're requesting.

Check that the Authorization header uses the Bearer prefix and that your token is current. If it has expired, obtain a new one from the Authentication API.

No. Ask a question in plain language and your AI client chooses the right tool, or combines several, to answer it.

Once the plugin is loaded, type / and select a skill, or make a natural-language request such as 'Compare these two funds' and your client selects the relevant skill automatically.

Your client decides which tools to call and how to combine the results, so the same question can produce different answers on different models. Morningstar recommends Claude models, or OpenAI GPT-5.1 or later, for reliable tool calling. Test your chosen model against your own workflows before rolling it out.

Every request starts by matching names and tickers to Morningstar identifiers, and many securities have similar names. Use tickers where possible in prompts and agent instructions.

Not in a single screen. Criteria combine with AND only, using equals, greater than and less than. For OR conditions, run separate screens and combine the results.

Only data points the lookup tool has mapped. Morningstar offers more than 1,300 data points (as of July 2026), but screening and retrieval by name are limited to the mapped subset.

Only the most recent reports and notes, typically available within 24 hours of publication. There is no archive, so research alone can't show how Morningstar's view has changed over time.

Up to 100 holdings per fund (the top 10 to 100), with portfolio weights. Only the largest holdings are returned, not the full portfolio.

The S&P 500, which is fixed. Non-US and multi-asset portfolios are therefore measured against a US equity benchmark.

Articles from 2022 onward.

Yes. Rate limits are determined by your contract.

Yes. The official integration for building custom agents powered by the MCP Server is in Microsoft Copilot Studio. Custom agents let you design targeted workflows for specific investment tasks that run consistently and repeatably.

Claude models are recommended, or OpenAI GPT-5.1 or newer. Whichever you choose, it should support structured tool calling, multi-step reasoning and reliable function invocation.

Go to + Add tool, select Model Context Protocol, find the official Morningstar MCP Server connector, enter the server URL, and configure authentication using OAuth 2.0 with dynamic discovery. When you send your first message, select Open Connection Manager and complete the sign-in.