Financial data MCP for AI agents

Why MCP changes financial-data integration

A traditional financial API makes the developer select endpoints, translate schemas, and write orchestration before an agent can use the data. MCP lets a compatible agent inspect tool names, descriptions, input schemas, and safety annotations at runtime. The agent can then choose a bounded tool for the task.

Unquant exposes one remote endpoint rather than a local package. A compatible host connects to https://unquant.ai/mcp, lists the tools, and calls only the dataset needed for the current question.

What an agent can research

The available catalog includes delayed or end-of-day quotes, daily price history, company profiles, company fundamentals, analyst ratings, earnings, ETF holdings, processed news summaries, macroeconomic events and indicators, and US congressional transaction disclosures.

• Compare companies using the latest reported fundamentals and state the reporting periods.
• Measure historical returns or event windows with normalized daily price bars.
• Combine scheduled macro events with company or market context.
• Review disclosed congressional transactions while preserving transaction and filing dates.

A reliable agent workflow

Start with the dataset catalog, select the narrowest matching tool, and preserve the response's dates, units, freshness, pagination, and limitations in the final answer. For long histories, reduce the data inside code execution instead of copying an entire series into the model context.

Financial information from Unquant is for research, not execution. Quotes are not real time, disclosures can arrive after the underlying transaction, and generated news or macro analysis can contain model error. Those boundaries are part of each public dataset contract.