What This Server Does
#Four MCP Tools
list_categories- the 8-category menu (also baked into the server instructions so an agent sees the full map at session start, no tool call needed).list_indicators_in_category- per-category indicator catalog (id, name, frequency, last date); optional compact mode.find_indicator- cross-category fuzzy substring search on indicator id and display name.v2_get_series- time series, one or more indicators, wide format (one row per date, one column per indicator). Date filters optional.
Eight Categories
- us_macro - FRED + Michigan: monetary, labor, inflation, GDP, sentiment, housing, financial conditions
- us_credit - delinquency + charge-offs (FRED, FFIEC), NCUA credit unions, NY Fed household debt, FDIC banks
- us_valuation - Shiller CAPE, Buffett indicator, nominal GDP, US Treasury yields
- us_insurance - Fed Z.1 invested-asset breakdown for Life and P&C insurers
- india_macro - RBI WSS, MoSPI (IIP, CPI, WPI, GDP), NSE Nifty 50 PE/PB/DivYield
- global_macro - OECD Business Confidence (37 countries), European sovereign yields
- markets - daily prices for 12 symbols (equity indices, US 10Y, DXY, gold, copper, WTI, Brent)
- currencies - daily FX (USDINR, USDJPY, USDKRW, USDTWD, USDTRY, USDPHP, USDVND)
Limits
- Max 10 indicators per
v2_get_seriescall - TSV output carries a
# metaline flagging any unknown or empty IDs - 60 req/min per IP
MCP Configuration
#Recommended - Streamable HTTP
https://api.tigzig.com/v2/mcp/http
Newer MCP spec (2025-03-26). Single endpoint, serverless-friendly. Try this first.
Fallback - SSE
https://api.tigzig.com/v2/mcp/sse
Legacy MCP transport (2024-11-05). Use if your client does not yet speak Streamable HTTP.
Adding to Claude.ai
- Open Claude.ai - Settings - Connectors
- Add custom connector, paste the Streamable HTTP URL
- If it does not connect, swap to the SSE URL
- Approve - the agent now has
list_categories,list_indicators_in_category,find_indicatorandv2_get_series
Same URLs work for Cursor, Continue, and any other MCP client. The MCP server identifies itself as "Tigzig V2 Unified Data Surface".
How It Works
#- FastAPI backend, same dataset that powers the TREMOR app charts
- Indicators stored in DuckDB, refreshed on the schedule of each upstream source (bulk downloads + replica 3x daily)
- MCP layer exposes four category-first tools over Streamable HTTP and SSE
- Open HTTP API mirrors the same data at
/v2/,/v2/categories,/v2/seriesfor non-MCP clients; bulk downloads at/v2/download/all - V1 (
/tremor/v1/) remains available for existing integrations; V2 is recommended for new ones - Cloudflare in front for rate limiting and WAF
Data Sources
#Indicators are pulled from primary publishers; the dictionary returns the upstream source and source series ID for each entry.
- US: FRED, NY Fed, NCUA, FDIC, Shiller
- India: RBI, MoSPI, NSE
- EU / Global: ECB, OECD
Sample indicator IDs: us_treasury_30y_yield, us_bank_delinq_cc, shiller_cape, india_credit_deposit_ratio.
Example Agent Prompts
#- "Pull the last 5 years of US 30-year Treasury yield and US credit card delinquencies. Are they correlated?"
- "Show me every macro early-warning indicator that is currently above its 2007 pre-crisis level."
- "List every Indian indicator you can pull and tell me which sources they come from."
- "Get Shiller CAPE for the last 145 years and tell me how today compares."