bcb-br-mcp is an MCP (Model Context Protocol) server designed to let AI assistants (such as Claude) directly query economic and financial time series data published by the Central Bank of Brazil (Banco Central do Brasil, BCB). It integrates three data sources: SGS time series (SGS/BCB), the Focus market expectations survey (provided via the Olinda OData API), and PTAX exchange rates.
Main capabilities: the server provides 17 tools (skills), 3 resources, and 3 prompts. Features include: querying time series values by code and date filter; retrieving the latest N values of any series; retrieving series metadata (frequency, source, etc.); providing a catalog of 135 source-verified popular economic indicators, organized by category; supporting keyword search (ignoring accents); retrieving the current values of key economic indicators; automatically handling long-period queries (the BCB API limits daily series to a 10-year window; the server automatically slices, fetches, and merges, supporting 15-year daily queries); frequency harmonization (series can be resampled to monthly, quarterly, or annual, and series that are already percentage changes are geometrically compounded, such as converting monthly IPCA to annual IPCA); calculating percentage changes over a period and statistics; comparing multiple series; querying market expectations from the Focus survey (mean, median, standard deviation, minimum, maximum, number of respondents for IPCA, GDP, exchange rate, etc., by monthly/quarterly/annual or trailing 12/24-month inflation, as well as Selic expectations by Copom meeting); querying official PTAX closing exchange rates (any currency published by the BCB, for a single day or a date range).
Usage: can be installed via Smithery (recommended), or used directly via an HTTP endpoint (no installation required), or configured in Claude Desktop via npx, and also supports global installation. ChatGPT deep research mode can use an MCP server that exposes only search and fetch, providing these two interfaces on top of the bcb_* tools.
Use cases: suitable for AI assistant users who need real-time, accurate Brazilian economic data, for example querying the current Selic rate, month-by-month IPCA for 2024, USD exchange rate changes over the past 12 months, market expectations for IPCA in 2027, Selic expectations for the next Copom meeting, Friday's EUR PTAX closing price, etc.
Limitations: data comes from BCB public APIs and requires no authentication. The SGS database has more than 18,000 time series, but the server's curated catalog has only 135, and search also covers the open data portal index (3,579 series). The BCB API imposes a 10-year window limit on daily series (HTTP 406), and dados/ultimos/N is capped at 20. There is no per-series metadata endpoint; frequency is inferred from observation intervals, and units of measurement are unavailable. Values calculated by the server are marked derived: true; values published by the BCB are returned as-is. Data is licensed under ODbL v1.0, and the code is MIT. Regarding privacy, no user data is logged.
