Market data is one of the most important layers in Charlie. It is also one of the best places to start.

Why? Because market data gives the assistant a trusted source for prices, instrument facts, market status, and news context. Without that layer, the assistant has two options: stay a bit generic, or look around the open web. For a financial institution, open-web search is usually not the right foundation. The information may be uncurated, inconsistently timestamped, commercially restricted, or simply not the source your teams have approved.

We understand that licensed news providers can be a significant cost driver, especially during an early Charlie pilot. That’s why we’re also exploring lower-cost retrieval options such as Tavily, where basic search can start at $0.008 per request and results can be limited to a controlled list of approved domains.

Start here

Market data is often easier to connect than people expect. In many cases, the first step is configuration: connect the approved provider, add the provider key or credentials, define which endpoints Charlie may use, and map the instrument identifiers.

That can already unlock useful first use cases.

Market data questions can include:

• “What is the current Nvidia price?”,

• “What was Meta’s closing price yesterday?”,

• “Is the market open right now?”
  
  News and context questions can include:

• “Why did this holding move today?”

• “Show me the latest approved news for the stocks I hold.”

This is also why market data pairs so well with a small amount of portfolio data. You do not need the full investing app connected on day one. If Charlie knows the user’s holdings, instrument identifiers, and basic portfolio context, market data can already make the answers much more relevant.

Keep it curated

The market-data layer should be boring in the best possible way: approved provider, clear source, clear timestamp, clear permissioning, and a trace of what was used.

That gives the model a much safer job. It does not have to invent market context. It can explain the market context returned by the tools.

For a pilot, this is a strong starting point: connect market data, add just enough portfolio context, and let Charlie answer a focused set of investor questions with real sources behind the response.