Kyomi MCP Server
Query your data warehouse from Claude Code, Cursor, and any MCP-compatible client.
Kyomi is an AI data intelligence platform that connects to your data warehouse and provides natural language querying, dashboards, automated monitoring, and accumulated business knowledge. The MCP server exposes all of this to your AI assistant.
Quick Start
Claude Code```bash
claude mcp add kyomi --transport http https://app.kyomi.ai/mcp
Restart Claude Code. You'll be prompted to authorize in your browser on first use.
### Cursor
1. Go to **Settings > Profile** in Kyomi
2. Click **Connect with Cursor**
3. Confirm the installation in Cursor
4. Authorize in your browser when prompted
### Manual Configuration
Add this to your MCP config file:```json
{
"mcpServers": {
"kyomi": {
"type": "http",
"url": "https://app.kyomi.ai/mcp"
}
}
}
Config file locations:
- Cursor: ~/.cursor/mcp.json
- Claude Code: ~/.claude/claude_desktop_config.json
Available Tools
Data Catalog
| Tool | Description |
|---|---|
search_catalog |
Semantic search across all connected datasources — "find customer revenue tables" |
get_table_info |
Detailed schema information for any table (columns, types, descriptions) |
list_datasources |
List all datasources connected to your workspace |
Query Execution
| Tool | Description |
|---|---|
query_datasource |
Run SQL queries against any connected data warehouse or database |
validate_sql |
Check query syntax before executing |
Knowledge Base
| Tool | Description |
|---|---|
search_learnings |
Access your team's accumulated knowledge about your data |
save_learning |
Capture new insights about tables, metrics, or business rules |
Dashboards
| Tool | Description |
|---|---|
search_dashboards |
Find existing dashboards by content or title |
get_dashboard_info |
View dashboard details and ChartML content |
create_dashboard |
Create new dashboards with ChartML visualizations |
Watches (Automated Monitoring)
| Tool | Description |
|---|---|
search_watches |
Find existing data monitors |
create_watch |
Set up automated data monitoring with alerts |
Website Analytics
| Tool | Description |
|---|---|
create_analytics_site |
Set up website analytics tracking and get the `` snippet |
list_analytics_sites |
View all analytics sites and their tracking snippets |
update_analytics_site |
Change site name or allowed domains |
delete_analytics_site |
Remove an analytics site and all its data |
Supported Datasources
Kyomi connects to 9 data platforms — all accessible through the same MCP tools:
| Platform | Type |
|---|---|
| BigQuery | Cloud Data Warehouse |
| Snowflake | Cloud Data Warehouse |
| PostgreSQL | Relational Database |
| MySQL | Relational Database |
| ClickHouse | Columnar Analytics DB |
| Redshift | Cloud Data Warehouse |
| Databricks | Lakehouse Platform |
| SQL Server | Relational Database |
| Azure Synapse | Cloud Analytics |
Example Usage
Once connected, ask your AI assistant:```
"Search for tables related to customer orders"
"What columns are in the sales.transactions table?"
"Run a query to get yesterday's revenue by product category"
"What does our team know about calculating MRR?"
"Create a dashboard showing weekly active users trend"
"Set up a watch to alert me if daily signups drop below 10"
```
Security
- Browser-based OAuth — authorize via your browser, tokens refresh automatically
- Workspace isolation — you only see your workspace's data
- Same permissions — MCP uses your existing Kyomi credentials and datasource access
- 20-row preview limit — AI queries return max 20 rows for context
- Your data stays in your warehouse — Kyomi queries directly, never copies or stores your data
For maximum security, deploy Kyomi Connect (open-source, Apache 2.0) inside your network so database credentials never leave your infrastructure.
Documentation
- Full MCP documentation
- Kyomi Connect (on-premise agent)
- Website
License
The Kyomi MCP server is a hosted service. This repository contains documentation and setup instructions.
Kyomi Connect, the on-premise database agent, is open-source under the Apache License 2.0.
