stat-gov-mcp
# stat-gov-mcp
Local MCP server for the **Polish Central Statistical Office (GUS) Bank Danych Lokalnych (BDL)** — the definitive public source for Polish socioeconomic statistics: population, prices, business demographics, unemployment, GDP, at country / voivodeship / powiat / gmina resolution.
Part of the [honest-mcp family](https://github.com/bartosz-kuc?tab=repositories) of small, auditable, local-first MCP servers.
## Why
BDL contains tens of thousands of time series about Poland. Its web UI ([bdl.stat.gov.pl](https://bdl.stat.gov.pl/)) is powerful but slow to navigate when you know exactly what you want. The public REST API is the fast path but has enough concepts (subjects → variables → units → data) to make manual use annoying. This server lets your AI find the right variable and unit and pull the data — one conversation, done.
Same trust model as the rest of the family: data flows only between your machine and GUS.
## Features
Five tools:
- `search_subjects` — browse or search the subject tree (e.g., "CENY", "LUDNOŚĆ")
- `search_variables` — find data series in a subject
- `search_units` — find territorial units (voivodeship, powiat, gmina) by name and/or level
- `get_unit_details` — full record for a unit ID
- `get_data` — pull actual values for a variable across chosen units and year range
## Data source
- Endpoint: [bdl.stat.gov.pl/api/v1](https://bdl.stat.gov.pl/api/v1/) — GUS BDL public REST API
- No API key required for the free tier (5 req/sec, ~5000 req/day)
- Higher-volume tier available with free registration; not needed for typical interactive use
## Requirements
- Python 3.10+
## Setup
```bash
git clone https://github.com/bartosz-kuc/honest-stat-gov-mcp.git
cd stat-gov-mcp
python3 -m venv venv
./venv/bin/pip install -r requirements.txt
```
Register with Claude Code:
```bash
claude mcp add stat-gov /absolute/path/to/venv/bin/python /absolute/path/to/server.py
```
Claude Desktop `claude_desktop_config.json`:
```json
{
"mcpServers": {
"stat-gov": {
"command": "/absolute/path/to/venv/bin/python",
"args": ["/absolute/path/to/server.py"]
}
}
}
```
## Example usage
> "How has the average monthly wage changed in Mazowieckie over the last 10 years?"
Three-step: `search_variables(name="przeciętne wynagrodzenie")` → note variable ID → `search_units(name="mazowieckie", level=2)` → note unit ID → `get_data(variable_id=..., unit_ids=[...], year_from=2015)`.
> "Which voivodeships had the highest unemployment in 2024?"
`search_variables(name="stopa bezrobocia")` → `get_data(variable_id=..., year_from=2024, year_to=2024)` — default returns all voivodeships.
## Data flow
```
Your AI client
↕ MCP stdio
This server (Python, on your machine)
↕ HTTPS
bdl.stat.gov.pl (GUS)
```
No cloud middle. No telemetry.
## Author
**Bartosz Kuć** — Warsaw-based developer, JDG owner running [skanfirmy.pl](https://skanfirmy.pl).
- GitHub: https://github.com/bartosz-kuc
- Email: firma@bartosza.pl
## Consulting
Available for consulting on Polish tax and business integrations (KSeF, GUS/NFZ/GIOŚ APIs, mBank data), MCP server design, and AI-assisted tooling for JDGs and small teams. See **[skanfirmy.pl/uslugi](https://skanfirmy.pl/uslugi)** for productized packages (audit 3k PLN, setup 8-15k PLN, retainer 2-4k PLN/mo), or reach out via email.
## License
MIT — see [LICENSE](LICENSE).
## Related
- Part of the honest-mcp family — see the [family index](https://github.com/bartosz-kuc?tab=repositories).
TDQS
Scored across 5 tools
Each tool targets a clearly distinct entity or action: subjects, variables, units, unit details, and data. The descriptions cross-reference IDs and usage patterns, making it easy for an agent to select the right tool.
Tool names follow a consistent verb_noun pattern: search_subjects, search_variables, search_units, get_unit_details, get_data. All use lowercase snake_case and the prefixes search_ and get_ clearly distinguish discovery from retrieval.
Five tools is well-scoped for a statistical data access server. Each tool covers a necessary part of the browsing and data retrieval workflow without unnecessary or redundant additions.
The core workflow is complete: discover subjects, search variables, locate territorial units, get unit details, and fetch data. Minor gaps exist such as detailed variable metadata or subject detail endpoints, but they are not essential for typical data retrieval.