How to Build an Executive Financial Dashboard With AI: MCP Connector or API Key?
Two ways to put live market data in front of an AI: a one-URL MCP connector for Claude, ChatGPT, and Grok, or an API key for code. Which fits an executive stock dashboard, the exact prompts to ask for a weekly briefing, and what the connector can and cannot do.
There are two ways to put live market data in front of an AI: an MCP connector you add to Claude, ChatGPT, or Grok (newly added!) by pasting one URL, or an API key your code calls directly. The rule for choosing: if a person is reading the output, use the MCP connector; if an agent or code is doing the work, use an API key. For most people that settles it, use the connector. The rest is knowing what to ask, and that is what this guide is for.
Which AI apps can connect today?
The hosted connector at https://app.sentisense.ai/mcp is verified in Claude, ChatGPT, and Grok. In each one you are looking for the same thing: the option to add a server by its URL. Claude and Grok call it Connectors; ChatGPT has renamed that area more than once, so look for Plugins. Step-by-step instructions for each app, with direct links to the right settings screens, are in the connector guide.
Local MCP clients that run on your own machine, like Claude Code, connect over loopback rather than a hosted callback. If you are wiring up a coding agent or anything running in your own process, that path works, and the API plus the agent skills are usually the better fit anyway.
MCP connector or API key?
| MCP connector | API key | |
|---|---|---|
| Setup | Paste one URL, sign in | Generate a key, wire up a client |
| Auth | Sign in with your account, nothing to copy | X-SentiSense-API-Key header |
| Surface | Curated read-only tools for chat | The full endpoint set, Python and Node SDKs |
| Best for | Ad-hoc questions, briefings, research | Agents, scheduled jobs, dashboards, products |
| Quota | Shares your account's monthly quota | Same quota, same tiers |
Both read the same data graph. The connector is the fastest way in; the API is the deeper one. And you do not have to write the code yourself: hand an agent like Claude Cowork, Codex, Grok Build, or Hermes Agent an API key plus the agent skills, and it can build and refresh the dashboard for you, with the full endpoint set and typed SDKs to work with.
What should you actually ask?
For a recurring executive view of a tracked list, four questions do most of the work. Ask them exactly like this, swapping in your own tickers:
- "Where is market mood today, and what changed this week?" A single fear-to-greed composite gives every other number context. A stock up 3% in a fearful tape is a different story than the same move in a euphoric one.
- "Rank NVDA, MSFT, and COST by SentiSense Score and flag the movers." Price, sentiment, and the Score for each ticker, so the list ranks itself instead of being read top to bottom.
- "What news explains this week's move in NVDA? Link the sources." Sentiment-tagged market-moving news, so the summary cites rather than asserts.
- "Which institutions added or trimmed COST last quarter, and did any analyst change a rating?" The slower signals that a weekly review is actually for.
Or ask for the whole thing at once. This is the prompt to save:
Give me an executive briefing on NVDA, MSFT, and COST. Start with overall market mood and what changed this week. Rank the three by SentiSense Score and flag anything that moved. Explain the movers with market-moving news and link the sources. Note any institutional buying or selling and any analyst rating changes. End with their next earnings dates.
Ask it every Monday morning and you have a weekly briefing. Reported financial statements and the earnings calendar are there when you want the numbers behind the narrative. Building instead of chatting? Each ask above maps to an endpoint; call them and render the dashboard yourself.
Quotes are real-time; sentiment, mood, and news annotations recompute on a schedule. That is the right cadence for an executive summary and the wrong one for an execution system.
Is it safe to connect?
- Read-only by design. It can look up data. It cannot trade, move money, or change anything on your account.
- You sign in, you can sign out. Access works through your SentiSense sign-in; no key or password is pasted into the chat app, and you can revoke the connection from your account at any time.
- It respects your tier. A free account gets free-tier depth. Connecting does not unlock more than you pay for.
- There is a named entity and published terms behind it. SentiSense is operated by Compass AI Data Services, LLC, and connector use is governed by the API Terms of Service. Data is provided for informational and educational purposes, not as investment advice.
Start with the connector, graduate to the key
The cheap experiment is the connector: paste one URL, ask the briefing prompt for a week, and find out which four numbers you actually reach for. Then build the dashboard against the API, knowing what belongs on it.
Connect SentiSense to Claude, ChatGPT, or Grok. No API key required.