Search, preview and download India’s public data catalogue from inside Claude, ChatGPT, Codex and any other assistant that supports remote MCP servers.
Dataful is a curated catalogue of India-focused public data: government statistics, economic indicators, and district-level series. The Dataful connector lets an AI assistant search that catalogue, preview datasets, and download the ones your account can access, all from inside a chat.
The Dataful MCP is built on the Model Context Protocol (MCP), so it works with any assistant that supports remote MCP servers.
Who it is for: researchers, journalists, analysts, students, and anyone who needs Indian public data without digging through ministry websites.
Install the Dataful plugin and connect your Dataful account.
Open Dataful in ChatGPT ↗Dataful is on its way to Claude’s connector directory. Until then, add it as a custom connector with the steps below.
Coming soonDataful is on its way to Muse. Until then, Muse Code users can add it with the steps below.
Coming soonYou need a Dataful account. Sign up for free at dataful.in.
Server URL:
https://mcp.dataful.in/mcp
Free, Pro and Max:
https://mcp.dataful.in/mcp and click Add.Free plans allow one custom connector.
Team and Enterprise: an Owner adds the connector first under Organization settings → Connectors → Add → Custom → Web. Each member then clicks Connect under Customize → Connectors.
To use it, click + in a chat, open Connectors, and turn on Dataful. On mobile, add the connector on the web or desktop first.
claude mcp add --transport http dataful https://mcp.dataful.in/mcp
Then run /mcp inside Claude Code and sign in.
On Business, Enterprise and Edu workspaces, an admin may need to enable Dataful first under Admin → Plugins.
ChatGPT asks you to confirm each action that writes, such as a download. Buying is not available in ChatGPT. See Buying a dataset.
codex mcp add dataful --url https://mcp.dataful.in/mcp codex mcp login dataful
In the IDE extension, open the gear menu → MCP servers → Add server, choose Streamable HTTP, enter the URL, restart the extension, and click Authenticate.
Add this to ~/.config/muse/settings.json:
{
"schema_version": 1,
"mcp_servers": {
"dataful": {
"transport": "streamable_http",
"url": "https://mcp.dataful.in/mcp",
"mode": "optional"
}
}
}Then run muse mcp login dataful and sign in. Run /mcp in a session to check the tools are listed.
The connector has nine tools. Seven only read data. Two do something on your account, and neither deletes or changes existing data.
| Tool | What it does | Changes anything? |
|---|---|---|
search | Finds datasets by topic in plain language. You can narrow by sector and years. | No |
list_datasets | Browses the full catalogue by sector, collection, access tier (free, high-frequency, premium) or district level. | No |
get_dataset | Shows one dataset in full: description, source, coverage, price, formats, columns, and a short preview of the first few rows. | No |
list_sectors | Lists sectors such as agriculture, health, and finance. | No |
list_collections | Lists curated collections of related datasets. | No |
list_downloads | Lists your own past downloads, newest first. | No |
get_order_status | Checks whether you have paid, and how long you can download for. | No |
download_dataset | Gets a short-lived download link for a dataset you can access. Records the download, and uses up allowance on a subscription. | Yes: records a download |
create_order | Starts buying a paid dataset and returns a payment page. Creating the order does not charge you. | Yes: creates an order |
Formats: tables download as CSV, XLSX or Parquet. Maps download as GeoJSON or Parquet.
Searching and previewing work for every dataset. Downloads depend on the dataset:
Things to know:
When a download is refused, the assistant tells you why. The common reasons:
| Prompt | Dataset it finds |
|---|---|
| “Analyse the air quality in Delhi over the last 3 years. Which months are the worst?” | Day wise, State wise Air Quality Index (AQI) of Major Cities and Towns in India |
| “How has monsoon rainfall in Kerala changed since 1901?” | Year, Month and IMD Sub Division-wise (State) Actual Rainfall Data Since 1901 |
| “Compare rural and urban inflation in India over the last five years.” | Consumer Price Index (CPI) - Master Data |
| “Which states grew their GST collections fastest after 2020?” | GST Collection: Year-, Month- and State-wise Gross Total GST Collection from States |
| “Which districts in Karnataka produce the most rice, and how has yield changed?” | District, Season and Crop wise Area, Production and Yield Statistics for Karnataka |
| “How will India's working-age population change by 2036? (free)” | Population Projection of India (2011 - 2036) |
| “How did the Nifty 50 perform during the 2020 crash and recovery?” | NSE Indices: Day-wise Open, Close, High, Low, Turnover Amount, and Shares Traded |
| “Which districts improved most on child stunting between NFHS-4 and NFHS-5?” | NFHS (4, 5 and 6): State- and District-wise Statistical Indicators |
| “Which datasets did I download last month?” | Your own download history |
Why can’t the assistant open the downloaded file?
download_dataset returns a link to the file on Dataful’s storage provider, Cloudflare R2. To analyse the file, the assistant’s code sandbox has to be allowed to reach that host. If it can’t, open the link yourself to download the file, then upload it to the chat. The link is valid for 5 minutes.
3653eb858375df80e1a202a906fea086.r2.cloudflarestorage.com
Can the assistant answer from the preview?
No. get_dataset shows only the first few rows, so it helps you decide whether a dataset is useful. For real answers, the assistant should download the file.
Search didn’t find a dataset I know exists.
Try more specific keywords, such as the source, geography or years. You can also ask the assistant to filter by sector, collection or district level.
A dataset says it has been replaced.
The dataset was withdrawn, and a newer one took its place. The assistant fetches the replacement automatically.
How do I disconnect?
Remove the connector in your assistant’s settings.
Email [email protected], or reach us through the Contact Us page.