A comprehensive guide to building a self-service analytics agent that integrates Slack, ClickHouse database, and AI capabilities. The tutorial demonstrates creating a Slack bot that processes natural language questions, converts them to SQL queries via PydanticAI and MCP (Model Context Protocol), executes them against ClickHouse, and returns human-readable results. The implementation uses Anthropic's Claude model, Socket Mode for Slack integration, and the ClickHouse MCP server for database connectivity, offering an alternative to traditional BI dashboards through conversational analytics.
Table of contents
What we’re building1. Setting Up Your Slack Bot2. Configuring and loading EnvVars2. Configuring PydanticAI to use mcp-clickhouse3. Configuring message handling4. Working on the message5. Running & testingConclusionSort: