What’s new here

Airbyte added a second mode alongside ELT: an Agent SDK (airbyte-agent-sdk) that wraps connectors as typed LLM tools with built-in retry, exception translation, and output-size guardrails. The SDK is a separate package whose own README lists 50+ third-party APIs, not the 600+ ELT connectors, and it lets a pydantic-ai or LangChain agent read live CRM or support data.

What it does

Airbyte moves data from sources (APIs, databases, files) to destinations (warehouses, lakes, databases, AI applications). It ships a catalog of 600+ connectors covering services like PostgreSQL, MySQL, S3, BigQuery, Redshift, and Snowflake.

You can deploy it yourself or use Airbyte Cloud. For custom sources, there is a no-code Connector Builder UI and a low-code CDK. Syncs can be orchestrated from Airflow, Dagster, Kestra, or the Airbyte API.

The Agent SDK is a separate install (uv pip install airbyte-agent-sdk) for teams building AI agents that need real-time access to business data rather than batch warehouse loads.

Who it’s for

Data engineers who need to centralize data from many sources without writing and maintaining custom sync scripts. Also relevant for AI teams that want to give agents access to live business data from SaaS APIs and databases without building bespoke connectors.

Try it

Deploy Airbyte locally or install the Agent SDK:

# Self-hosted
# See https://docs.airbyte.com/quickstart/deploy-airbyte
 
# Agent SDK
uv pip install airbyte-agent-sdk

How mature is it

22,201 stars, 5,397 forks, 100+ contributors, 100+ releases with the latest at v2.0.0 (October 2025). Active development with 100+ commits in the past 90 days and last push on 2026-10-10. The license is a mix of MIT and Elastic License v2 (ELv2) depending on the component; check the repo’s LICENSE file and FAQ for specifics.