IBGE Brasil MCP

IBGE Brasil MCP is an independent, remotely operated Model Context Protocol server from Sidney Bissoli for access to official Brazilian geographic, demographic, and statistical data. The official product site at https://ibge.sidneybissoli.com describes 23 tools that query official IBGE APIs live and return a provenance block with the table, period, and reproducible URL for each response. The documented remote endpoint is https://ibge.sidneybissoli.com/mcp and uses Streamable HTTP. The provider repository at https://github.com/SidneyBissoli/ibge-br-mcp is public, TypeScript-based, and MIT licensed. Smithery lists sidneybissoli/ibge-br-mcp as a verified and deployed remote server. The product site, repository, provider tutorial, and Smithery deployment information identify the same durable product and endpoint. This catalog stores no credentials, tokens, API keys, or personal data.

Purpose and data domains

According to the provider, the server connects compatible AI clients to public IBGE data sources instead of supplying numbers from a model's training snapshot. Documented areas include states, municipalities, locality codes, neighboring municipalities, and geographic meshes. Other tools query SIDRA tables, metadata, census datasets, research series, and economic or social indicators. The repository names population, gross domestic product, prices, unemployment, education, and income among the possible domains. A health tool returns indicators served through IBGE's SIDRA system. Tools for names, economic-activity classification, international country information, news, and release calendars extend the scope. Search and fetch tools for ChatGPT Deep Research also implement the documented contract for catalog search and citable retrieval.

The interface is described by the provider as read-oriented. Results can contain raw observations, metadata, or comparisons and distributions computed by the server. For derived statistics, users must distinguish values directly returned by IBGE from values calculated by the server from raw observations. The provenance block is intended to make that origin visible; it does not replace expert review of the table, definition, unit, territorial level, or reference period. Data quality depends on choosing the appropriate IBGE table, on source completeness, and on whether query parameters correctly represent the question. A plausible number is not proof of currentness or methodological fitness.

Authorization, privacy, and data sharing

The product site and repository describe the hosted endpoint as requiring no key. That does not mean that every use is free of access restrictions, rate limits, or operational rules. Clients must check current provider and IBGE documentation, protect the expected host with TLS, and allow only the official endpoint. If access requirements change, secrets must remain only in the client's secure authentication mechanism. They must never be put in prompts, seeders, logs, or the catalog.

IBGE publishes official statistical and geospatial information for the public, government, and research communities. Even so, queries, geographic references, usage logs, or combinations of results can enable sensitive inferences. Before use, review data minimization, logging, retention, deletion, storage region, subprocessors, and possible sharing with the connected model provider. A local MCP client does not make this remotely operated service local processing: the query and response reach the endpoint and may then be processed in the context of the selected AI model. Public official data does not remove privacy duties for additional user or business information.

Access restrictions, write access, and security

The published product description identifies read-only tools. In the checked documentation there is no intended write access to IBGE datasets, so the server should not be understood as a way to alter official data. Write access can nevertheless arise outside this service when an AI client transfers results into files, databases, reports, or external systems. Such actions require separate authorization, least privilege, and human approval. The client must prevent a tool response from granting itself new network or file permissions.

Remote responses, table metadata, and free-form questions are untrusted input and may contain prompt injection. Use endpoint allowlisting, TLS, timeouts, logging without secrets, minimal data sharing, and citation review. Respect possible access, rate, and availability limits of the public IBGE APIs and the hosting operation. The IBGE portal at https://www.ibge.gov.br is the authoritative place for institutional context and original data; this MCP server is an access layer, not an official IBGE instance. The E-E-A-T basis of this entry is, according to the provider, the product site, repository, documentation, and Smithery deployment information. These sources establish product identity and documented scope, not the correctness of every model response. The repository's nine GitHub stars on September 9, 2026 are only a time-bound popularity signal and not evidence of quality or security.

An MCP client with Streamable HTTP support and network access to the official remote endpoint. Reliable answers require suitable IBGE queries, source review, and secure model configuration.

Configure the official remote endpoint https://ibge.sidneybissoli.com/mcp in the MCP client using current provider guidance. Before production use, verify the endpoint, tool schema, data sharing, and possible access restrictions.

https://ibge.sidneybissoli.com/mcp

According to the provider, the hosted endpoint requires no key. Verify current authentication and access requirements with the provider; never store secrets in the catalog or prompts.

The tools are read-only according to the provider. Respect rate limits, availability, and possible restrictions of the hosting service and IBGE APIs. External client write actions require separate approval.

Queries reach the remote endpoint and responses contain source and table data. Data may then be processed or shared in the connected model context; minimize it, review logging, and check retention.

Prompt injection in questions or responses, wrong table selection, outdated or derived values, sensitive inferences, and unintended client writes. Mitigate with allowlisting, TLS, least privilege, source review, and human approval.