datagovin-mcp

An MCP server that gives Claude (or any MCP client) natural-language access to India's Open Government Data platform, data.gov.in — 100,000+ public datasets covering air quality, agriculture, health, fuel prices, census, education, rainfall and more.

The server ships no data of its own. It's a thin, well-behaved wrapper: every query is a live call to data.gov.in using your own free API key. What it adds is the missing layer — discovery and clean querying — so a language model can go from "what's the AQI in Delhi right now?" to real rows without you ever hunting for a dataset UUID by hand.

Why this exists

data.gov.in has an enormous catalog but no good full-text search API: the normal workflow is to browse the website and copy a dataset's resource ID off its "API" button. That's a poor fit for an LLM. This server closes the gap with a curated, verified index you can search by keyword, plus tools that fetch a dataset's schema and pull filtered rows live. It also absorbs the upstream API's rough edges (case-sensitive filters, occasional CSV responses, a last-page pagination quirk) so the model doesn't have to.

Related MCP server: bharatlas-mcp

Tools

Setup

1. Get a free API key. Register at data.gov.in and generate a key from your profile page. It's free with no published rate limit.

2. Install.

git clone https://github.com/<your-username>/datagovin-mcp.git
cd datagovin-mcp
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt

3. Set your key.

cp .env.example .env
# edit .env and paste your key, or just export it:
export DATA_GOV_IN_API_KEY="your_key_here"

4. Connect it to Claude Desktop. Add this to your MCP client config (claude_desktop_config.json):

{
  "mcpServers": {
    "datagovin": {
      "command": "python",
      "args": ["/absolute/path/to/datagovin-mcp/server.py"],
      "env": { "DATA_GOV_IN_API_KEY": "your_key_here" }
    }
  }
}

Restart the client and the four tools appear.

Example conversation

You: What datasets do you have on air quality? Claude: (calls search_datasets("air quality")) → Real-time Air Quality Index by station.

You: What fields does it have? Claude: (calls get_dataset_info(...)) → country, state, city, station, pollutant_id, pollutant_avg, last_update…

You: Show me PM2.5 readings in Delhi. Claude: (calls query_dataset(resource_id, filters={"city": "Delhi", "pollutant_id": "PM2.5"})) → live rows.

Growing the catalog

The catalog ships seeded with a verified dataset and is meant to grow. To add any dataset, copy its resource ID from the "API" button on its data.gov.in page and run:

python scripts/add_dataset.py <resource_id> --sector Agriculture --keywords "wheat,crop,production"

The script fetches the dataset's real title and schema live and appends a verified entry to data/catalog.json. --sector and --keywords are optional.

Notes on the upstream API

A few data.gov.in behaviours this server handles for you:

• Filter field names are case-sensitive (filters[State] ≠ filters[state]) and this is undocumented. Always use the exact field id from get_dataset_info.

• Some legacy datasets return CSV regardless of format=json; the client detects this via Content-Type and parses it anyway.

• Last-page pagination can return an empty records array with status: ok; a zero-count result with a positive total means you've paged past the end.

• Max ~100 rows per request — page through larger sets with offset.

Project layout

datagovin-mcp/
├── server.py              # MCP entry point (stdio); defines the four tools
├── datagovin/
│   ├── client.py          # async data.gov.in API wrapper (quirk handling)
│   └── catalog.py         # curated catalog load + keyword search
├── data/catalog.json      # verified, extensible dataset index
├── scripts/add_dataset.py # enrich the catalog by resource ID
├── requirements.txt
└── pyproject.toml

License

MIT