What this enables for you
++ Whether you are prototyping a policy assistant or embedding data in + an internal tool, the server is designed so the model can + look up facts instead of guessing numbers or mixing + units. +
+-
+
-
+
Country and regional insight
++ Compare economies, track SDG-related series, and see whether an + indicator actually covers the places you care about. +
+
+ -
+
Trends over time
++ Request years or periods you need—then reason over consistent + time series instead of a vague “recent” snapshot. +
+
+ -
+
Definitions you can cite
++ Pull methodology, limitations, and field meanings so answers can + explain what the number measures—not only its value. +
+
+ -
+
Charts when it helps
++ Turn suitable series into visualization specs (for example + Vega-Lite) so assistants can show trends, not only tables. +
+
+
+ Questions your assistant can support +
++ The tools are built for natural, analyst-style questions. Your agent + resolves countries to standard codes, picks indicators, applies + filters (sex, age, urbanization, etc.), and checks coverage before + answering. +
+-
+
- + How has GDP per capita changed in Kenya over the last two decades? + Search, confirm coverage, then return a time series. + +
- + Which indicators in WDI relate to female labor force participation, + and do they include Bangladesh? + Discovery plus metadata and disaggregation checks. + +
- + What is the official definition and source for this indicator—and + what footnotes should I mention? + Metadata retrieval for transparency and reporting. + +
- + Plot a line chart comparing access to electricity in three + countries since 2010. + Data first, then visualization specs when the series fits. + +
Charts and visualizations
++ Beyond tables, the server can help agents produce chart-ready output + for time series and comparisons—so “show me the trend” becomes a + structured spec your client can render. +
++ Typical flow: fetch data with clear columns (for + example time period and observed value), then ask for a + visualization specification. Chart type hints such as line, bar, + or scatter are mapped to supported views; the agent can discover + valid options and constraints via dedicated tools. +
+-
+
- + Line and area — trends over years or periods + +
- + Bar — discrete periods or categories when + appropriate + +
- + Scatter / point — relationships when the data + supports it + +
+ Specs are produced in a standard chart grammar (Vega-Lite) so + compatible hosts can render them consistently. If you operate a + separate chart rendering service, you can point the server at it + with environment configuration for integration. +
+
+ Related tools:
+ data360_get_viz_spec,
+ data360_get_supported_chart_types
+
+ Technical reference +
++ The sections below are for developers wiring an MCP + client, IDE, or custom agent: capabilities summary, tool and resource + tables, install steps, and connection URLs. +
+Server capabilities
++ Implementation details behind the user-facing outcomes above—search, + metadata, series retrieval, codes, and LLM-oriented resources. +
+-
+
-
+
Smart indicator discovery
++ Search hundreds of indicators with enriched metadata and optional + country coverage checks. +
+
+ -
+
Rich metadata
++ Methodology, definitions, limitations, and statistical concepts + on demand. +
+
+ -
+
Reliable time series
++ Historical data with filters for country, period, sex, age, + urbanization, and more. +
+
+ -
+
LLM-oriented resources
++ System prompts, codelists, and chain-of-thought guidance for + chatbot integration. +
+
+ -
+
Agent-friendly design
++ Composable tools and guardrails that reduce common data retrieval + mistakes. +
+
+
MCP tools
++ Use these tools in sequence: search, then fetch data with the right + codes and filters. +
+| Tool | +Description | +
|---|---|
data360_search_indicators |
+
+ Search with enriched metadata; use
+ required_country for coverage. Returns
+ covers_country, latest_data,
+ dimensions.
+ |
+
data360_get_data |
+ + Fetch data points with filters (country, time period, SEX, + AGE, etc.). + | +
data360_get_metadata |
+
+ Indicator metadata; use select_fields to limit
+ payload.
+ |
+
data360_get_disaggregation |
+ + Available filter values (countries, years, dimensions) per + indicator. + | +
data360_find_codelist_value |
+
+ Resolve names to codes (e.g. “Kenya” →
+ KEN, “female” → F).
+ |
+
data360_list_indicators |
+ List all indicators for a database. | +
data360_get_viz_spec |
+ + Build a chart specification from retrieved data (e.g. Vega-Lite). + | +
data360_get_supported_chart_types |
+ List supported chart types and data requirements. | +
data360_get_data_api_url |
+ Low-level: direct Data360 data API URL helper. | +
+ Recommended agent workflow +
+1. Search → data360_search_indicators(query, required_country="Kenya") + Returns: covers_country, latest_data, dimensions per indicator + +2. Get data → data360_get_data(database_id, indicator_id, filters) + Use REF_AREA from search; add time period filters+
MCP resources
+
+ Subscribe to these URIs for prompts and reference data. For chatbots,
+ include data360://system-prompt in your system prompt.
+
| Resource | +Description | +
|---|---|
data360://system-prompt |
+ Chain-of-thought guidance for chatbot integration | +
data360://databases |
+ Available databases (WB_WDI, WB_SSGD, …) | +
data360://codelists |
+ Codelist reference (REF_AREA, SEX, AGE, …) | +
data360://metadata-fields |
+ Field mapping for smart question routing | +
data360://data-filters |
+ Available filters and usage guidance | +
data360://search-usage |
+ Search examples and best practices | +
Get started
+
+ Requires Python 3.11+ and
+ uv (recommended) or
+ pip. Copy .env.example to .env
+ and adjust API URL or port as needed.
+
Install & run
+# Clone and install (uv) +git clone https://github.com/worldbank/data360-mcp.git +cd data360-mcp +uv sync +cp .env.example .env +uv run poe serve +# Server: http://localhost:8000/mcp+ +
Configuration
+| Variable | +Description | +Default | +
|---|---|---|
DATA360_API_BASE_URL |
+ Data360 API base URL | +https://data360api.worldbank.org |
+
MCP_PORT |
+ MCP server port | +8000 |
+
MCP_TRANSPORT |
+ http or sse |
+ http |
+
MCP_CHARTS_API_URL |
+ Optional chart rendering API URL | +— | +
+ Try the bundled demo:
+ uv run scripts/llm_mcp_demo.py (use
+ DEBUG=true for verbose logs).
+
Connect your agent
+
+ Point your MCP client at the local HTTP or SSE endpoint. Use
+ host.docker.internal instead of
+ localhost from containers.
+
Endpoints
+| Transport | +http (default) or sse |
+
|---|---|
| URL (HTTP) | +http://localhost:8000/mcp |
+
| URL (SSE) | +
+ http://localhost:8021/sse (example port)
+ |
+
Databases (examples)
+
+ All Data360 databases are supported; list indicators with
+ data360_list_indicators.
+
-
+
- WB_WDI — World Development Indicators +
- + WB_SSGD — Social Sustainability and Global + Database + +