dlt & dltHub

dltHub is the company behind dlt, an Apache 2.0 open-source Python library for building data pipelines with code and AI coding agents. Its legal entity is ScaleVector GmbH in Berlin, with a US entity and hubs in San Francisco, Montreal and Amsterdam. It targets engineers who want ingestion as code instead of closed connector catalogues.

With dlt, a pipeline is a Python script that extracts from APIs, databases or files, infers and evolves schemas, normalises nested data and loads it incrementally into warehouses and lakes. The site reports 6M+ PyPI downloads per month and 16,000+ companies running dlt in production. The managed dltHub platform adds a runtime, observability, data quality checks, team workspaces and background agents that monitor jobs. Case studies include PostHog, Harness, Artsy and Flatiron Health.

• Python-native pipelines: Pipelines are plain Python code that runs anywhere Python runs.
• Schema inference and evolution: Schemas are inferred, normalised and updated automatically.
• Incremental loading: Built-in incremental and merge loading to warehouses and lakes.
• Agent-assisted development: Context for 10,100+ sources helps AI agents build connectors.
• Managed runtime: dltHub deploys, schedules and monitors pipelines with data quality checks.
• Background agents: Public-preview agents inspect jobs and keep pipelines healthy.

A Vietnamese tech startup can load data from local payment, e-commerce and marketing APIs into BigQuery using Python scripts generated with AI agents. A bank's data team can replace hand-written ingestion scripts with versioned dlt pipelines. An analytics agency can run many client pipelines on the managed dltHub runtime.

The open-source library is free for commercial use, so teams can start without licence costs and avoid per-row connector pricing. Pipelines live in code and version control, and AI agents can generate connectors from more than 10,100 documented sources. The vendor claims migrations from legacy vendors run 90% faster. The managed platform adds monitoring and agents for teams that do not want to run infrastructure.

Data teams are shifting from closed, volume-priced connector services to code-first ingestion that AI coding assistants can write and maintain. Engineering-led companies in Southeast Asia, where budgets for per-row ETL pricing are tight, are natural adopters of open-source pipeline tooling.

Legal entity ScaleVector GmbH, Rosenthaler Str. 36, Berlin (HRB 241116 B), with dltHub Inc. in the US. Co-founded by Matthaus Krzykowski, Marcin Rudolf, Anna Hoffmann and Adrian Brudaru. Backed by Bessemer Venture Partners, which led an $8M round, plus Foundation Capital and Dig Ventures.