Skip to main content

Datasets

The Knowledgator MCP server provides comprehensive dataset management, synthetic data generation, and stateless upload/download tools for autonomous AI agents and coding assistants.


Available Dataset Tools​

  • Upload Dataset — Upload validated dataset records or JSON strings to create a new dataset (upload_dataset).
  • Download Dataset — Download dataset records directly in memory via storage redirect (download_dataset).
  • Generate Dataset Plan — Generate a structured schema and sample plan from a prompt (generate_dataset_plan).
  • Generate Dataset — Start an asynchronous synthetic dataset generation pipeline (start_dataset_generation).
  • List Datasets — Query and filter platform datasets with pagination (list_datasets).
  • Get Dataset — Retrieve detailed metadata, schema, and sample counts for a dataset (get_dataset_details).
  • Dataset Remediation & Skills — Automated 5D quality diagnostics, semantic auditing, and deterministic repair executed by Knowledgator's on-platform Managed Agent.

Example Workflow​

Agents can execute a complete data lifecycle end-to-end:

  1. Plan & Generate: Call generate_dataset_plan to formulate taxonomy, then trigger start_dataset_generation.
  2. Verify: Call get_dataset_details with the dataset ID to check status (valid, ready).
  3. Download: Call download_dataset to inspect samples or verify label distributions.
  4. Curate & Upload: Curate, clean, or augment samples and call upload_dataset to publish an updated version.