🏦 FRED + Treasury Auctions Macro Data MCP for AI Agents

One MCP that answers macro questions end-to-end — FRED economic series + US Treasury yield curve + Treasury auction calendar in a single tool-call. Stop wiring three separate API integrations into your LLM trading agent — point Claude / Cursor / Windsurf at this MCP server and your agent can answer "what's the 10Y, what's CPI YoY, what's this week's auction calendar, what did the last 30Y bond auction stop out at?" in one prompt.

Why this MCP server exists

Macro AI agents — LLM-powered trader assistants, fixed-income copilots, hedge-fund research bots — are exploding as a buyer category. But every macro-aware agent needs the same three data feeds: FRED for the macro indicators, the Treasury par yield curve for the rates picture, and the Treasury auction calendar for the supply picture. Today those are three different APIs with three different auth schemes, three different rate limits, and three different data shapes.

This MCP server collapses them into five tools:

Tool	What it returns
get_fred_series(series_id, start_date, end_date)	Any FRED series — CPIAUCSL, GDP, UNRATE, FEDFUNDS, DGS10, T10Y2Y, M2SL, PAYEMS, ... — with series metadata + observations
get_treasury_yield_curve(date)	Full 1mo / 2mo / 3mo / 4mo / 6mo / 1y / 2y / 3y / 5y / 7y / 10y / 20y / 30y par yield curve for a date, plus the 10y-2y spread
get_upcoming_treasury_auctions(security_type, days_ahead)	Bill / Note / Bond / TIPS / FRN / CMB auction calendar — auction date, issue date, maturity, offering amount, CUSIP
get_auction_results(auction_id_or_recent, security_type, days_back, limit)	Historical auction results — stop-out high yield, bid-to-cover ratio, total tendered/accepted, primary-dealer/direct/indirect allotments
get_macro_dashboard()	One-call snapshot — latest 10Y yield, full curve, CPI YoY, headline unemployment, effective fed funds rate, last 5 auction results

Buyer personas

• Fixed-income portfolio managers running an LLM copilot for daily desk briefings
• Macro hedge fund analysts who want the same data point + auction surprise scoring in 10 seconds, not 30 minutes
• AI trading agents — ChatGPT/Claude agents that need real-time macro context to answer customer questions
• LLM-powered trader assistants wired into broker dashboards, where the macro picture is one of N tools
• Quant data ops prototyping a Bloomberg-replacement stack on top of public data

Quick start — wire it into Claude Desktop / Cursor / Windsurf

This is a Standby MCP server. Once published it lives at https://nexgendata--fred-treasury-macro-data-mcp.apify.actor/mcp and responds to the standard streamable-HTTP MCP transport. Add this to your client config:

{
  "mcpServers": {
    "fred-treasury-macro": {
      "url": "https://nexgendata--fred-treasury-macro-data-mcp.apify.actor/mcp",
      "transport": "http",
      "headers": {
        "Authorization": "Bearer YOUR_APIFY_TOKEN"
      }
    }
  }
}

(Cursor / Windsurf use the same shape; n8n's MCP node accepts a URL + bearer too.)

To use the FRED tools you need a free FRED API key — grab one in 30 seconds at https://fredaccount.stlouisfed.org/apikeys and paste it into the actor input fred_api_key. The Treasury tools (get_treasury_yield_curve, get_upcoming_treasury_auctions, get_auction_results) work without any key.

Example agent prompts

Once wired in, your agent can answer questions like:

• "What's the current Treasury yield curve and the 10y-2y spread?" → get_treasury_yield_curve
• "Give me CPI year-over-year as of the latest print" → get_fred_series(series_id="CPIAUCSL", start_date="2025-01-01")
• "What Treasury auctions are coming up in the next 14 days?" → get_upcoming_treasury_auctions(days_ahead=14)
• "How did the last 10Y note auction stop out — was there a tail?" → get_auction_results(security_type="Note", days_back=21)
• "What's the macro picture right now?" → get_macro_dashboard() (one call, full snapshot)

What get_macro_dashboard() returns (sample)

{
  "as_of": "2026-05-30",
  "yield_curve": {
    "record_date": "2026-05-30",
    "curve_pct": { "1M": 4.92, "3M": 4.85, "2Y": 4.32, "5Y": 4.18, "10Y": 4.41, "30Y": 4.68 },
    "spread_10y2y_pct": 0.09
  },
  "cpi": { "series_id": "CPIAUCSL", "latest": {"date":"2026-04-01","value":318.4}, "yoy_pct": 2.7 },
  "unemployment": { "series_id": "UNRATE", "latest": {"date":"2026-04-01","value":4.1} },
  "fed_funds_rate": { "series_id": "FEDFUNDS", "latest": {"date":"2026-04-01","value":4.33} },
  "ten_year_yield": { "series_id": "DGS10", "latest": {"date":"2026-05-29","value":4.41} },
  "recent_auctions": [
    {"cusip":"912797PN0","security_type":"Bill","security_term":"4-Week","auction_date":"2026-05-28","high_yield_pct":"4.825","bid_to_cover_ratio":"2.94"}
  ]
}

That's one tool-call. The same agent prompt against three separate APIs is six round-trips and a custom join.

Data sources

• FRED — Federal Reserve Economic Data, St. Louis Fed (https://api.stlouisfed.org/fred/). 800K+ economic series. Free with API key. License: CC0 / public domain.
• US Treasury Fiscal Data — https://fiscaldata.treasury.gov/. Daily par yield curves, upcoming auctions, securities auction results. No key required. License: public domain.

How it works under the hood

• Apify Standby HTTP transport — the actor stays warm and answers MCP requests in milliseconds once spun up
• Five tools, all async, all charge tool-call on successful invocation
• get_macro_dashboard() fans out 6 parallel upstream calls (1 yield curve + 1 auctions + 4 FRED series) using asyncio.gather, so the dashboard latency is dominated by the slowest single upstream
• Graceful error envelopes — when FRED rate-limits or Treasury is briefly down, the tool returns a structured error rather than crashing the MCP session
• 240-second no-client watchdog exits cleanly if no MCP client ever connects (so you don't pay for an idle container)

🔗 Related Actors

Pair this with these other NexGenData MCP servers and data feeds:

• 🤖 Finance MCP Server — AI-Native Stock & Market Data — Yahoo Finance + Finviz + CoinGecko quotes for LLM trading agents. The equity-side complement to this macro/fixed-income MCP.
• 📊 Regulatory Filings MCP Server — SEC EDGAR + regulator filings for the same LLM agent stack. Pair with this macro MCP for top-down + bottom-up coverage.

About NexGenData

NexGenData publishes 250+ buyer-intent actors covering SEC filings, regulatory enforcement, real-estate, lead generation, competitive intelligence, fundamentals across 30+ exchanges, and a growing fleet of MCP servers purpose-built for AI agents. All actors are pay-per-event. Browse the full catalog at https://apify.com/nexgendata?fpr=2ayu9b.

🏠 Home: thenextgennexus.com 📦 Full catalog: apify.com/nexgendata