VayuChat MCP

VayuChat MCP enables natural language data analysis for air quality datasets using predefined MCP functions. You can ask questions, run analyses, and generate visualizations without writing code, all through a consistent, reliable interface that executes specific, well-defined operations.

How to use

You interact with the MCP server through an MCP client. Start by connecting to the server and asking questions in natural language. Your queries are interpreted to call predefined MCP functions in a reliable sequence, returning structured results and visualizations. You can explore available data, compare metrics across cities, analyze weekday versus weekend patterns, examine funding trends, and build city profiles with a few simple prompts.

Common tasks you can perform include data exploration (listing tables, describing tables, filtering with queries), analysis (weekday vs weekend comparisons, city comparisons, correlations, funding breakdowns), and visualizations (bar charts, time series, histograms, trend charts, hourly patterns). Use straightforward questions like “What tables are available?”, “Show me the air quality data for Delhi”, or “Plot weekday vs weekend PM2.5” to get immediate results.

How to install

Prerequisites you need before installation are Python and a Python package manager. You also need a runtime capable of executing MCP servers, such as uv. Make sure you have access to the terminal or command prompt.

Install the MCP package using the recommended commands. Run these in your terminal:

# Using uv
uv pip install -e .

# Or with pip
pip install -e .

Configuring and running the MCP server

If you want to run the MCP server as an integrated service, you can use the following configuration to start the server through the runtime. This config snippet defines the MCP server with the command and arguments you should run.

{
  "mcpServers": {
    "vayuchat": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/vayuchat-mcp", "vayuchat-mcp"]
    }
  }
}

Running and testing

Start the MCP server using the configured command, then connect your MCP client to the server. Once connected, begin with simple exploratory questions to verify the setup. For example, list available tables, describe a table, and run a basic query to confirm responses and formatting match your expectations.

Examples of typical prompts you can try include: “What tables are available?”, “Show me the funding table”, “Describe the air quality data”, and “Plot weekday versus weekend PM2.5”.

Available tools

list_tables

Show available tables in the loaded datasets.

show_table

Display data from a specific table.

describe_table

Provide detailed statistics and summaries for a table.

query_table

Filter and query table data using a pandas-like syntax.

compare_weekday_weekend

Compare metrics between weekdays and weekends.

compare_cities

Contrast metrics across multiple cities.

analyze_correlation

Compute correlations between selected metrics.

analyze_funding

Breakdown and analyze funding data by city and year.

get_city_profile

Generate a comprehensive profile for a chosen city.

plot_comparison

Create bar or box charts to compare metrics.

plot_time_series

Visualize data as time series plots.

plot_weekday_weekend

Plot weekday vs weekend comparisons.

plot_funding_trend

Visualize funding trends across years.

plot_hourly_pattern

Show hourly patterns for chosen metrics.