Switching between your AI assistant and AstroBee to check data structures or run queries breaks your flow. With MCP (Model Context Protocol), your AI assistant connects directly to your data layers: so you can explore schemas, run SQL, and get insights without leaving Claude Code, Claude Desktop, or VS Code.
MCP is an open standard that lets AI assistants securely connect to external tools and data sources. AstroBee’s MCP server gives your assistant read access to your data layers and the ability to execute queries on your behalf.
Always visible in the left sidebar, the MCP page is where you manage these connections.
Coming soon: An Active Connections table will let you monitor all connected AI assistants, view session details, and track connection status in real time.
Quick setup
Claude Code
Claude Desktop
Other MCP Clients
- Run this command in your terminal:
claude mcp add astrobee --transport http https://agent.production.bumblebee.kurtosis.io/mcp
- In Claude Code, type
/mcp and select “Authenticate”
- Sign in with your AstroBee account in the browser
Learn more about MCP in Claude Code.
- Open Settings → Connectors → Add custom connector
- Enter this URL:
https://agent.production.bumblebee.kurtosis.io/mcp
- Click Connect and sign in with your AstroBee account
Use this JSON configuration for any MCP-compatible client:{
"astrobee": {
"type": "http",
"url": "https://agent.production.bumblebee.kurtosis.io/mcp"
}
}
For editing data layers, use the AstroBee web interface rather than MCP tools. The web interface provides optimized workflows that prevent token overhead and maintain data layer consistency. If you have a specific use case that requires programmatic data layer modifications, contact us at [email protected]. Server & model discovery
| Tool | Description |
|---|
server-status | Check server health and connection statistics |
list-ontologies | List all data layers (semantic models) you have access to |
list-source-models | Get raw data schemas linked to a business model |
Entity & property exploration
| Tool | Description |
|---|
list-entities | List all entities in a model with metadata (optionally with row counts) |
list-properties | Get all properties for a specific entity |
describe-property | Get detailed info about a specific property (type, domain/measure) |
list-relationships | View how entities are connected in a model |
Statistics & data quality
| Tool | Description |
|---|
list-entity-stats | Get entity-wide metrics (row count, property counts, quality scores) |
list-property-stats | Get property statistics (uniqueness, null %, distinct counts, samples) |
SQL execution
| Tool | Description |
|---|
get-datasource-type | Get the database engine (BigQuery, Postgres, Snowflake) for correct SQL syntax |
validate-sql | Check if a SQL query is syntactically correct without executing |
execute-sql | Run a SQL query and get results (up to 1000 rows) |
Next steps
