Skip to main content
Glama
README.md
# open-data-mcp

[![markatdataandbeyond/open-data-mcp MCP server](https://glama.ai/mcp/servers/markatdataandbeyond/open-data-mcp/badges/score.svg)](https://glama.ai/mcp/servers/markatdataandbeyond/open-data-mcp)
[![Listed on mcpservers.org](https://mcpservers.org/badge.svg)](https://mcpservers.org/servers/markatdataandbeyond/open-data-mcp)

Ten tools I built to get facts from official open data, not guesses. This repository is an **open-source MCP server** (stdio). Each tool calls the official source API directly from your machine — no Apify token required for the local server.

Optional env keys (never commit them):

- `OPENALEX_API_KEY` — enables OpenAlex in scholar-search (free key from openalex.org)
- `NCBI_API_KEY` — higher PubMed rate limits (free from NCBI)

## Try in 30 seconds

```bash
npm ci
node scripts/demo-mcp-call.mjs dataandbeyond--eu-company-verify '{"queries":["DK24256790"]}'
```

Live output (2026-10-31, Europe/Copenhagen): `DK24256790 → NOVO NORDISK A/S`, VIES `valid=true`, LEI `549300DAQ1CVT6CXN342`. No API key required for this call.

## Tools

| Tool | Official source |
|---|---|
| `dataandbeyond--eu-company-verify` | EU VIES + GLEIF |
| `dataandbeyond--dk-statistics` | Statistics Denmark (Statbank) |
| `dataandbeyond--scholar-search` | Crossref, arXiv (+ OpenAlex with key) |
| `dataandbeyond--pubmed-search` | NCBI PubMed E-utilities |
| `dataandbeyond--eu-tenders` | TED API v3 |
| `dataandbeyond--france-company-search` | Recherche d'entreprises (INSEE/DINUM) |
| `dataandbeyond--nordic-company-registry` | Brønnøysund + PRH/YTJ |
| `dataandbeyond--clinical-trials` | ClinicalTrials.gov API v2 |
| `dataandbeyond--vin-decoder-recalls` | NHTSA vPIC + Recalls |
| `dataandbeyond--weather-forecast` | MET Norway (+ Nominatim geocode) |

## Run locally

```bash
npm ci
node src/index.js
```

Point your MCP client at the stdio command `node /path/to/open-data-mcp/src/index.js` (or the Docker image built from the Dockerfile in this repo). `tools/list` works with no environment variables.

Docker:

```bash
docker build -t open-data-mcp .
docker run --rm -i open-data-mcp
```

## Hosted version

The same ten tools are also available as **pay-per-event Apify Actors** if you want a hosted endpoint instead of running the server yourself:

| Actor | Store | Price (USD) |
|---|---|---|
| [eu-company-verify](https://apify.com/dataandbeyond/eu-company-verify) | VIES + GLEIF | 0.004 per company verified |
| [dk-statistics](https://apify.com/dataandbeyond/dk-statistics) | Statbank | 0.003 per table described + 0.002 per 100 rows |
| [scholar-search](https://apify.com/dataandbeyond/scholar-search) | papers | 0.002 per paper |
| [pubmed-search](https://apify.com/dataandbeyond/pubmed-search) | PubMed | 0.002 per article |
| [eu-tenders](https://apify.com/dataandbeyond/eu-tenders) | TED | 0.003 per tender notice |
| [france-company-search](https://apify.com/dataandbeyond/france-company-search) | Sirene | 0.003 per company |
| [nordic-company-registry](https://apify.com/dataandbeyond/nordic-company-registry) | NO+FI | 0.003 per company |
| [clinical-trials](https://apify.com/dataandbeyond/clinical-trials) | ClinicalTrials.gov | 0.003 per trial |
| [vin-decoder-recalls](https://apify.com/dataandbeyond/vin-decoder-recalls) | NHTSA | 0.001 per VIN / 0.003 with recalls |
| [weather-forecast](https://apify.com/dataandbeyond/weather-forecast) | MET Norway | 0.001 per forecast |

Hosted MCP URL (requires your Apify token):

```
https://mcp.apify.com/?tools=dataandbeyond/eu-company-verify,dataandbeyond/dk-statistics,dataandbeyond/scholar-search,dataandbeyond/pubmed-search,dataandbeyond/eu-tenders,dataandbeyond/france-company-search,dataandbeyond/nordic-company-registry,dataandbeyond/clinical-trials,dataandbeyond/vin-decoder-recalls,dataandbeyond/weather-forecast
```

Client config examples for the hosted path: [examples/cursor.json](examples/cursor.json), [examples/claude-desktop.json](examples/claude-desktop.json), [examples/vscode.json](examples/vscode.json).

## Demo

Short terminal recordings of the local stdio server calling official open data APIs. No API keys required for these demos.

| Tool | What you see | GIF | Screenshot |
|---|---|---|---|
| `eu-company-verify` | EU VIES check for DK24256790 → NOVO NORDISK A/S | ![eu-company-verify demo](assets/demo/eu-company-verify.gif) | [PNG](assets/demo/eu-company-verify.png) |
| `vin-decoder-recalls` | NHTSA decode of a Honda Accord VIN + recall count | ![vin-decoder-recalls demo](assets/demo/vin-decoder-recalls.gif) | [PNG](assets/demo/vin-decoder-recalls.png) |
| `scholar-search` | Crossref/arXiv papers for "knowledge graph embedding" | ![scholar-search demo](assets/demo/scholar-search.gif) | [PNG](assets/demo/scholar-search.png) |

Replay a cast with [asciinema](https://asciinema.org/): `asciinema play assets/demo/eu-company-verify.cast`. Re-record with `bash scripts/record-local-demos.sh`.

## License

MIT. See [LICENSE](LICENSE).

TDQS

A3.5/5.0

Scored across 10 tools

Disambiguation4/5

Each tool targets a clearly distinct data source and domain (statistics, papers, tenders, trials, vehicles, weather), so most selections are unambiguous. However, three tools overlap on company lookup (nordic-company-registry, eu-company-verify, france-company-search), and an agent handling an unspecified company could hesitate between them; only country/jurisdiction clues in the descriptions disambiguate.

Naming Consistency4/5

All names use a consistent namespace prefix plus kebab-case domain labels, so the format is predictable. There is minor stylistic drift between verb-led names (scholar-search, pubmed-search, eu-company-verify) and pure noun labels (dk-statistics, eu-tenders, weather-forecast), and country qualifiers are applied inconsistently (nordic vs france vs eu).

Tool Count5/5

Ten tools, each mapping to one distinct official API/data domain, is well within the ideal 3-15 range. Every tool clearly earns its place and none appears redundant (the apparent company overlap is jurisdictional, not duplicative).

Completeness3/5

Each tool is a self-contained connector covering its own source reasonably, but as an 'open data' surface the collection is an arbitrary sample — no Eurostat/World Bank/other-country company registries, no cross-source discovery tool, and no way to know what domains exist beyond the ten listed. The surface has no generic fallback, so agents hitting an uncovered dataset or country are at a dead end.

Maintenance

ActivityMaintained
ResponsivenessNo recent community issues