PaywallDataAgent

MCP server extension for querying Paywall data from Azure Data Explorer (Kusto). Provides a suite of tools for paywall telemetry analysis — funnel metrics, purchase data, A/B test results, and more — directly from VS Code via the Model Context Protocol (MCP).

Features

Query paywall telemetry data from Azure Data Explorer (Kusto)
Funnel analysis: sessions → UI shown → purchase clicked → purchase success
Purchase metrics, conversion rates, and entry point breakdowns
A/B test experiment lookup and flight-based analysis
Table schema exploration for Android and iOS paywall tables
Export query results to formatted Excel files
Built-in KQL validation to prevent dangerous or malformed queries
Platform-aware context with sample queries for Android and iOS

Available Tools

Tool	Description
`get_sample_queries`	Returns schema, table descriptions, clusters, and sample KQL queries filtered by platform. Call this first before writing any query. Covers funnel analysis, purchase counts, conversion rates, session data, entry points, A/B tests, error analysis, and general telemetry.
`validate_kql`	Validates a KQL query before execution — checks for blocked commands and syntax errors. Always call before `query_kusto_live`.
`query_kusto_live`	Executes a validated KQL query on Azure Data Explorer. Returns row count, column names, a result file path, and a preview of the first 20 rows. Use for funnel analysis, purchase metrics, session counts, A/B test results, etc.
`kusto_list_tables`	Lists all tables in the Kusto database for a given platform (`android` or `ios`).
`kusto_get_schema`	Returns column names and types for a specific Kusto table.
`export_to_excel`	Exports Kusto query results to a formatted `.xlsx` file in `~/Downloads`. Pass the `result_file` path from `query_kusto_live`.
`search_experiment`	Fuzzy-searches for an A/B test experiment by partial name. Returns top matches with IDs.
`get_experiment_allocations`	Returns full experiment details: audiences, progression steps, running status, flight allocation IDs, and bucket ranges. Use after `search_experiment` to get treatment/control IDs for funnel queries.

Typical Workflows

Paywall Funnel Analysis

get_sample_queries — load context and sample patterns
Ask filtering questions (platform, date range, app, device, entry point)
validate_kql — validate the constructed query
query_kusto_live — execute and get results
export_to_excel — (optional) save results to Excel

A/B Test Experiment Analysis

search_experiment — find the experiment by name
get_experiment_allocations — get treatment/control allocation IDs
get_sample_queries — load context for building the flight-based query
validate_kql → query_kusto_live — execute the query

Requirements

Python 3.10+
VPN access to Kusto clusters
Azure CLI (az login) for authentication

Getting Started

Install the extension
Connect to VPN
Run az login in your terminal
The MCP server starts automatically on activation