What it does

The dltHub AI Harness is a collection of toolkits for AI coding assistants. The page describes dlt as an open-source Python library for loading data from APIs and databases into a warehouse or lakehouse, and dltHub as the paid platform on top of it. Each toolkit bundles skills, commands, rules, agents and an MCP server behind a defined workflow, so the assistant follows a fixed sequence of steps instead of generating ad-hoc code. Toolkits cover guided setup, REST API and SQL database ingestion, loading files from disk or cloud storage, data exploration with dashboards, transformations, data quality checks, performance tuning and deployment to the dltHub Platform. A local MCP server lets the assistant list tables, preview data, run SQL, read schemas and view secrets with values masked, and a remote dltHub context lets it search more than 9,700 REST API source definitions for verified connectors before writing code. It is tested with Claude Code, Cursor and Codex and is installed with the dlthub-init command or the dlthub ai CLI. A Claude Code marketplace plugin is offered as early access. Several toolkits are marked as needing sign-up, and the harness is released under its own dltHub AI Harness License.

Capabilities

Guided workflows for REST API, SQL database and file-based ingestion pipelines
Local MCP server for inspecting tables and schemas without pasting output into chat
Search across 9,700+ REST API source definitions before code is written
Toolkits for data exploration, transformations, data quality and performance tuning
Installs into Claude Code, Cursor or Codex with the dlthub ai CLI

When to use it

You build data pipelines with dlt and want your coding assistant to follow a defined workflow
You want the assistant to inspect loaded data and schemas directly through MCP
You want a guided first run from raw data to a dashboard