Connect to RisingWave using a Postgres-compatible client.
Create a source with CREATE SOURCE specifying the connector type (kafka, kinesis, pulsar, etc.), topic or stream name, and format (JSON, Avro, Protobuf, etc.).
Optionally specify schema using WITH and FORMAT/ENCODE clauses; for Avro, provide the schema registry URL.
Create a materialized view with CREATE MATERIALIZED VIEW AS SELECT ... FROM <source> to begin continuous computation.
Query the materialized view like a regular table; results reflect the latest computed state.
Monitor progress with system tables such as rw_materialized_views and rw_source_backfill_progress.
Known gotchas
RisingWave sources are shared definitions; dropping a source used by multiple materialized views requires dropping all dependent views first.
The connector properties and format clause syntax evolve across releases; always match your client version to the cluster version in production.
Materialized views in RisingWave are maintained incrementally but are not cost-free — complex multi-join views can introduce significant memory and CPU overhead.
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