MCP Usage

With the MCP server running, you can ask AI assistants natural language questions. The server provides ~14K tokens of instructions that teach the LLM how to route questions to the right tools across 40+ APIs.

How It Works

When the MCP server connects to a client (VS Code, Claude Desktop, Cursor), it sends:

1. Instructions (~14K tokens) — module descriptions, a cross-reference routing table, and analysis rules. This stays in the model's context for the entire session.
2. 300+ tool definitions — each with name, description, and parameter schema.

The routing table tells the LLM which specific tools + parameters to combine for each question type. For example, when you ask about the deficit, the model reads:

DEBT/DEFICIT → FRED(fred_series_data with GDP, FYFSGDA188S) + Treasury(query_fiscal_data with debt_to_penny, avg_interest_rates) + World Bank(wb_indicator with GC.DOD.TOTL.GD.ZS)

...and knows to call those 3 tools with those specific parameters, then cross-reference the results.

Example Prompts

Economic

"What's the current state of the U.S. economy? Show me GDP, unemployment, inflation, and interest rates."

Tools: fred_series_data (GDP, UNRATE, CPIAUCSL, FEDFUNDS)

"Compare Biden and Trump's economic performance using the same metrics."

Tools: fred_series_data, treasury_query (debt), fr_executive_orders

"Give me a complete economic and health profile for Texas."

Tools: bea_gdp_by_state, census_query, cdc_places_health, usaspending_by_state, hud_fair_market_rents

Legislative

"What happened with the Inflation Reduction Act? Who sponsored it, how did the vote break down by party?"

Tools: congress_bill_details, congress_senate_votes, congress_bill_actions, fred_series_data (before/after), lobbying_search

"Who's lobbying on AI regulation and how much are they spending?"

Tools: lobbying_search, lobbying_detail, congress_search_bills

Health

"Show me the adverse event profile for Ozempic including clinical trials, FDA reports, and pharma payments to doctors."

Tools: fda_drug_events, fda_drug_counts, fda_drug_labels, fda_approved_drugs, fda_drug_shortages, clinical_trials_search, open_payments_search, nih_search_projects

Financial

"Show me which banking PACs gave money to members of the Senate Banking Committee, and how those members voted on banking deregulation."

Tools: congress_search_members, fec_search_committees, fec_committee_disbursements, congress_senate_votes, lobbying_search

"Which banks have the most consumer complaints?"

Tools: cfpb_search_complaints, cfpb_complaint_aggregations, cfpb_complaint_trends

Tips for Better Prompts

1. Be specific about time ranges — "GDP since 2020" beats "GDP trends"
2. Ask for cross-references — "...and compare internationally" triggers World Bank data
3. Name the data source — "Using FRED data, show me..." helps pick the right tool
4. Ask for context — "What was the debt-to-GDP ratio?" pulls GDP alongside debt