Finding a development indicator should be simple. In practice, the data are spread across many organisations, APIs and catalogue systems. Similar indicator names can also hide important differences in definitions, units and methods.

Impact Engines' Development Indicators MCP gives compatible AI assistants one consistent way to search and retrieve official statistics from sources including the World Bank, WHO, UNICEF, FAOSTAT, ILOSTAT, IMF, the UN SDG database, UNESCO, OECD and ADB.

What is an MCP?

Model Context Protocol (MCP) is a standard way to connect an AI assistant to external tools and data. It is like giving the assistant a specialised set of search and retrieval functions that it can call while answering a question.

Without this connection, an assistant may rely on general web search or information learned during training. With the Development Indicators MCP, it can search participating data catalogues, read their metadata and request current observations in a consistent, structured format. The assistant does not need to learn a different API for every organisation, and the user does not need to know in advance which organisation publishes the best indicator.

What it does

  • Searches official indicator catalogues.
  • Inspects definitions, units and source metadata.
  • Resolves country names to standard codes.
  • Retrieves observations directly from publishers.
  • Compares source-qualified series without silently treating them as equivalent.
  • Retains citations, provider codes, dimensions and status flags.

It does not invent estimates or merge similar-looking indicators. The aim is to help an agent reach a traceable answer while preserving the statistical context needed to assess it.

Why it is needed

General web search is useful for finding reports, but it is a weak interface for repeatable statistical work. An AI assistant may select the wrong series, omit a unit, confuse national and modelled estimates, or provide a number that is hard to reproduce.

This MCP connects the assistant to official data APIs and returns structured results with their provenance. It reduces the time spent learning different API formats without removing the need to check definitions and comparability.

Set it up

The hosted endpoint is https://indicatorsmcp.impactengines.ai/mcp. No command line or API key is required.

ChatGPT desktop app / Codex

The easiest way is to paste this directly into a Codex chat:

Install the Development Indicators MCP from https://indicatorsmcp.impactengines.ai/mcp

If that does not work, open Settings > Plugins, select the MCPs tab, choose Add > Add MCP server, enter Development Indicators as the name, select Streamable HTTP, enter the endpoint above, and save.

Claude

The easiest way is to paste the same instruction directly into a Claude chat:

Install the Development Indicators MCP from https://indicatorsmcp.impactengines.ai/mcp

If that does not work, open Settings > Connectors, select Add, enter Development Indicators as the name and the endpoint as the MCP server URL, select Continue, then choose No sign-in.

Example prompts

Using the Development Indicators MCP, find annual GDP growth for Ghana, Kenya and Uganda from 2019 to 2024. Choose an appropriate official series and include its source, definition, units and citations.

Using the Development Indicators MCP, find official maize production data for Uganda and Kenya from 2020 to 2023. Explain which source and series you selected, and preserve units, flags and source links.

Using the Development Indicators MCP, compare the latest suitable unemployment indicators for Ghana and Kenya. Explain the source selected and any differences in frequency, definition or dimensions before comparing the values.

Search the available official sources for maternal mortality indicators. Show the strongest matches and explain why similarly named series may not be directly comparable.

Codex or Claude may select the MCP automatically when a question clearly asks for development statistics. For more reliable results, begin with Using the Development Indicators MCP.... You do not normally need to name a data source; finding an appropriate official source is one of the MCP's main jobs.

Open source

The complete source code is available under the permissive MIT License. You can inspect the adapters and safeguards, contribute improvements, fork the project, or host your own copy with Docker or Cloudflare Workers.

GitHub: https://github.com/lbruce-wq/development-data-mcp

The hosted service is currently a public beta. Feedback on indicator discovery, metadata quality, source coverage and real research workflows is welcome.