AI access via MCP

Connect Claude — or any MCP client — to every table and series on this site.

@reserve-bank-innovation-hub/dbie-mcp is an MCP server that lets AI assistants search, browse and fetch RBI economic data from this site. It is a thin, read-only layer over the same static JSON the pages render — no credentials, no writes, nothing to host.

Set it up

With Claude Code, one command:

claude mcp add dbie -- npx -y @reserve-bank-innovation-hub/dbie-mcp

For any other MCP client (Claude Desktop, Cursor, VS Code, …), add this to its MCP configuration:

{
    "mcpServers" : {
        "dbie" : {
            "command" : "npx",
            "args"    : [ "-y", "@reserve-bank-innovation-hub/dbie-mcp" ]
        }
    }
}

Node 20 or newer is the only requirement — npx fetches and runs the package on demand.

What the assistant can do

Tool	What it does
`search_tables`	Full-text search across every table and series, typo-tolerant
`list_tables`	Browse the catalogue — sections, then the tables in each
`get_series`	Observations for a series, sliced by date range and columns, with units
`get_table`	Any table’s full JSON payload, long arrays truncated to protect context

A typical exchange: ask “how have India’s forex reserves moved over the last five years?” — the assistant calls search_tables, picks the weekly reserves series, then pulls exactly that window with get_series.

Pointing at another deployment

Data comes from DBIE_BASE_URL, defaulting to https://dbie.rbihub.in. To read from a different environment:

{
    "command" : "npx",
    "args"    : [ "-y", "@reserve-bank-innovation-hub/dbie-mcp" ],
    "env"     : { "DBIE_BASE_URL" : "https://dev.dbie.rbihub.in" }
}

Data vintage— responses reflect the deployment’s last scrape of the DBIE portal, not live data. Every response carries its source URL, and series responses carry as_of — the latest observation period — so assistants can (and should) disclose how current the numbers are.