# Data Airbyte Connector State and Incremental Syncs
Status: public
Confidence: medium (0.685) (verified)
Last verified: 2026-06-03
Generation: ai_structured


## TL;DR

Airbyte connector state lets agents reason about incremental sync progress, checkpointing, and what data may need replay after a failure.

## Core Explanation

Incremental pipelines need a memory of what they already copied. In connector-based ELT, that memory is often a cursor or state message that moves forward as records are read and written.

Agents troubleshooting a sync should capture the stream name, cursor field, last emitted state, last committed destination offset, job attempt, source ordering guarantees, and whether a full refresh or partial backfill occurred. Without that context, retrying a connector can miss late updates, duplicate records, or falsely appear complete.

## Source-Mapped Facts

- Airbyte protocol documentation describes sources and destinations exchanging serialized JSON messages over standard input and output. ([source](https://docs.airbyte.com/platform/understanding-airbyte/airbyte-protocol))
- Airbyte protocol documentation describes state messages as a mechanism for checkpointing replication. ([source](https://docs.airbyte.com/platform/understanding-airbyte/airbyte-protocol))
- Airbyte incremental sync documentation says an incremental sync retrieves only new or modified records by using a cursor field. ([source](https://docs.airbyte.com/platform/connector-development/connector-builder-ui/incremental-sync))

## Further Reading

- [Airbyte Protocol](https://docs.airbyte.com/platform/understanding-airbyte/airbyte-protocol)
- [Airbyte Incremental Sync](https://docs.airbyte.com/platform/connector-development/connector-builder-ui/incremental-sync)