Define an asset with a DailyPartitionsDefinition (or other partition type) and implement the asset function to accept the partition_key context argument
Create a sensor using the @asset_sensor or @multi_asset_sensor decorator that monitors a source asset or an external trigger for new partition completions
In the sensor's evaluation function, yield RunRequest objects with partition_key set to the newly available partition to trigger incremental asset materializations
Configure the sensor's minimum_interval_seconds to control polling frequency and avoid overloading the Dagster scheduler
Verify in the Dagster UI's partitions view that only the requested partitions are materialized and that the partition status reflects success or failure per partition key
Known gotchas
Sensors run in a separate process on a polling interval; if a sensor emits many RunRequests in a single tick, Dagster queues them but does not guarantee they run immediately or in order — use cursor state in the sensor to avoid re-emitting already-triggered partitions
Partition definitions are immutable after an asset is deployed; changing the partition definition (e.g., changing the start date) requires wiping the partition history or creating a new asset key
Multi-asset sensors that monitor multiple upstream assets fire when any of the monitored assets materialize; the sensor logic must explicitly check which partitions of which assets are new to avoid triggering downstream runs prematurely
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp