Kyomi

What is Kyomi?

Kyomi is an MCP server that enables AI assistants to query your data warehouse, access dashboards, automated monitoring, and accumulated business knowledge. It integrates with Claude Code, Cursor, and any MCP-compatible client.

How to use Kyomi?

Install via Claude Code with claude mcp add kyomi --transport http https://app.kyomi.ai/mcp, or connect via Cursor settings. Manual configuration is also possible. Restart the client and authorize in your browser on first use.

Key features of Kyomi

• Semantic search across data catalog
• Run SQL queries and validate syntax
• Access team's accumulated business knowledge
• Create and search dashboards with ChartML
• Set up automated data monitoring with alerts
• Manage website analytics tracking

Use cases of Kyomi

• Find tables related to customer orders
• Get column details for a specific table
• Run revenue queries by product category
• Retrieve business knowledge about metrics like MRR
• Create dashboards showing weekly active users trend

FAQ from Kyomi

What data platforms does Kyomi support?

Kyomi connects to BigQuery, Snowflake, PostgreSQL, MySQL, ClickHouse, Redshift, Databricks, SQL Server, and Azure Synapse.

How does authentication work?

Browser-based OAuth is used; tokens refresh automatically. Workspace isolation and your existing Kyomi credentials apply.

Is my data copied or stored by Kyomi?

No. Kyomi queries your warehouse directly and never copies or stores your data. A 20-row preview limit is enforced for AI queries.

Can I deploy Kyomi on-premise for maximum security?

Yes, Kyomi Connect (open-source, Apache 2.0) can be deployed inside your network so database credentials never leave your infrastructure.

Where is the MCP configuration file stored?

For Cursor: ~/.cursor/mcp.json; for Claude Code: ~/.claude/claude_desktop_config.json.