Connecting ChatGPT, Claude, Cursor or another AI tool to Massive uses the Massive MCP server (MCP is the Model Context Protocol, an open standard for connecting AI tools to outside systems). The address is the same for each of the supported tools; what differs is where each tool asks you to enter it, so the setup pages linked below carry the current steps for each one. A free Massive Basic account is enough to start. The table further down sorts common problems by the side they come from: the AI tool's own plan or settings, or the plan on the Massive account you signed in with.
Applies to
- Plans: every Massive plan, including the free Basic plan. Some AI tools also need a paid plan or a setting of their own; each tool's setup page says which.
- Endpoints or feeds: the hosted Massive MCP server at https://mcp.massive.com/, which calls the Massive REST API for you.
- Asset classes: the ones your Massive plan includes.
- AI tools with a setup page: Claude Code, Claude Desktop, Cursor, ChatGPT, Grok, Codex, GitHub Copilot and Gemini CLI (the command-line tool, not the Gemini web app).
How to connect an AI tool to Massive
Connecting an AI tool to Massive starts from the Massive setup page for that tool:
The steps are the same shape for each tool:
- Open the setup page for your AI tool from the list above.
- Add https://mcp.massive.com/ where that page says: a custom connector in the chat apps, an MCP server entry in the coding tools.
- When the AI tool opens your browser (for some tools, the first time you ask for Massive data), sign in to your Massive account and allow access. There is no API key to copy or paste.
- Test the connection with a prompt that names it, such as "Use the Massive connector to show me the dividends AAPL has paid this year."
Menu names in the AI tools change, so follow the current setup page for your tool rather than a menu path from an older guide. Each setup page ends with a troubleshooting section for that tool.
If you are building your own AI application rather than using one of these tools, the AI Tools quickstart also covers calling the Massive REST API directly, with an API key, through function calling.
Which side the problem is on
When a Massive connector fails, the symptom tells you which side the problem is on, the AI tool or your Massive account:
| What you see | Which side | What to do |
|---|---|---|
| Your AI tool has no option to add a custom connector or MCP server | Your AI tool: its plan, a setting such as developer mode, or a workspace admin | Check the requirements on your tool's setup page. Some tools need a paid plan or developer mode, and in an organization or team workspace an admin may have to allow or add the connector first. |
| The browser sign-in does not finish, or Massive's sign-in page shows an error such as invalid_client or "The requested OAuth 2.0 Client does not exist" | The sign-in step, before your Massive plan is checked | Reconnect Massive from your AI tool and finish the browser sign-in, as the troubleshooting section of your tool's setup page describes. If the same error page comes back, start a chat at massive.com/contact with the error text and the time you saw it. |
| The AI tool answers with numbers that look made up, or does not mention Massive | Your AI tool: it answered from its own knowledge instead of calling Massive | Start the prompt with "use the Massive connector to" and name the data you want. |
| An access-denied error | Your Massive plan: the asset class is not in it | Check your subscriptions in the dashboard. See How can I upgrade my Massive account? |
| Data is delayed when you expected real time | Your Massive account: the connector has the same access as the account you signed in with | Check what you would check for an API key: the plan tier for that asset class, and for real-time data the exchange agreements. See I paid for real-time Massive data, so why is it still delayed or refused? and How can I upgrade my Massive account? |
Why the connector has the same access as your Massive account
The Massive connector has the same access as your Massive account because you sign in with that account instead of pasting a key, and the MCP server makes its calls through the same live REST API that programmatic integrations use. The AI Tools quickstart says access through the MCP server mirrors your Massive entitlements: you can query any asset class your plan includes, and a request for one it does not include, such as options data without an Options plan, returns an access-denied error.
Limits of the hosted MCP server
The hosted Massive MCP server currently provides 64 MB of working storage per session and up to 10 concurrent workspaces, which is roughly 10 simultaneous chats, according to the AI Tools quickstart. The quickstart calls that ample for typical analysis. If you routinely download very large datasets, the self-hosted server below may suit you better.
Self-hosted MCP server for large datasets
The self-hosted Massive MCP server runs in your own environment instead of on Massive's servers, so its data workspace is on your own machine: it stores results as in-memory tables, runs SQL queries against them and applies built-in financial functions such as Greeks, returns and technical indicators. Install it with uv or pip from the official repository, github.com/massive-com/mcp_massive. Unlike the hosted server, which you sign in to in the browser, the self-hosted server needs your Massive API key, set in the MASSIVE_API_KEY environment variable.
