Databox MCP exposes a set of tools (API methods) that the AI or client can invoke to perform various actions. These tools cover everything from data ingestion, querying, to management of datasets and metrics. They give your AI assistant full programmatic control over your Databox environment. Below is a list of the key tools available and what they do:
| Tool Name | Description |
|---|---|
ask_genie | Runs AI-powered analysis on a dataset, translating natural questions into structured queries to produce insights. |
create_data_source | Creates a new data source entry to organize datasets within Databox. |
create_dataset | Creates a new dataset in Databox with a defined or auto-inferred schema. |
delete_data_source | Deletes a data source along with all datasets it contains. |
delete_dataset | Deletes a specified dataset from Databox. |
get_current_datetime | Get the current date and time to resolve relative date expressions like "last month" or "yesterday". |
get_dataset_ingestions | Shows ingestion history for a dataset to verify when data was last added. |
get_ingestion | Get detailed information for a specific ingestion event, including processing status and record counts. |
ingest_data | Ingests records into a Databox dataset, creating or updating it automatically. |
list_accounts | Lists all Databox accounts accessible with the current credentials. |
list_data_source_datasets | Lists all datasets contained within a specified data source. |
list_data_sources | Retrieves all data sources available in the current account. |
list_metrics | Lists all metrics available for a data source with their keys, names, descriptions, and dimensions. |
load_metric_data | Load data for a metric over a date range with optional dimension breakdowns and time series granulation. |
Note: Additional tools may be added over time as the MCP server continues to evolve. In the current release, there are 14 primary tools as listed above, offering comprehensive control over both datasets and existing Databox metrics. All tool calls use the MCP's JSON-RPC protocol — usually, as a developer you won't call these directly, but the AI agent will use them under the hood to fulfill your requests.