Deliver results somewhere
Configure a writable endpoint and approved policy, save and run a bounded pipeline, then verify delivered rows or files in the receiving system.
Use a destination when a verified workflow result needs to reach another system. Start with a small input and a target whose contents you can inspect after execution.
Steps
- In Plugins, explore capabilities filtered by Writes. Install and configure a plugin that supports the target system.
- In Data Endpoints, create a destination endpoint using that capability. Set the target and schema hints. Check access and confirm the connector's supported write behavior.
- Configure the applicable approved write policy and actor permissions. Append, replace, and upsert have different identity and deletion semantics. Source writes also require their own enabled access and grants.
- In Pipeline Studio, add or configure a destination node referencing the endpoint and the upstream data pool. Inspect its write mode and target.
- Validate and run configuration preflight. Preflight does not perform a test write or guarantee that external effects can be rolled back.
- Save version, then Run a small bounded input. Inspect the step result and verify the actual rows or files in the receiving system.
Where a human decision is required for a source mutation, use a published governed action with a required approval policy. Its request, approval, execution, and receipt are distinct steps. Do not assume selecting node approval metadata creates that action workflow.
External writes can be irreversible. Before retrying a partial or uncertain write, reconcile target state and inspect its receipt or recovery operation.
See Write policies and Read your first run.