Databases & data stores

Data systems have first-class adapters with their full operation set:

• SQL databases: PostgreSQL, MySQL, MariaDB, SQL Server, Oracle and ClickHouse. Write any SQL in the statement editor; record values bind as :name parameters instead of being pasted into the SQL, so a record value cannot change the statement. The SQL you write runs as written. Compatible services (Amazon Aurora and RDS, Redshift, CockroachDB, Neon, Supabase, TimescaleDB and more) use the same adapters; search the node library for the product name.
• Other databases through ODBC: DB2, IBM i, Informix, Teradata, SAP HANA, SAP ASE, Progress and any database with an ODBC driver. These connections run on an On-Prem Worker that has the database's ODBC driver installed; the connection form lists the drivers and data sources each Worker found. Write SQL in the statement editor with :name parameters, insert rows, and browse schemas, tables and columns. Large text and binary values are read in pieces, and stopping a run stops the running statement.
• IBM MQ: put, get and browse messages with their properties and descriptor fields, binary bodies included, check queue depth and list queues, on any queue manager (including z/OS and IBM i). Install IBM's MQ client on the On-Prem Worker that runs the step; the Worker's page shows whether it found one. Messages you get leave the queue only after the flow has them all, so a failed or stopped run returns them to the queue.
• Documents: MongoDB (find, insert, update, replace, delete, aggregation, count, distinct, collection and index administration, and a command operation), Azure Cosmos DB and Google Firestore.
• Key-value and cache: Redis and compatibles (Valkey, KeyDB, ElastiCache) and Memcached.
• Search: Elasticsearch, Typesense, Meilisearch.
• Vector stores: Qdrant, Weaviate, Pinecone; pgvector through PostgreSQL.
• Time series: InfluxDB v2.
• Warehouses: Snowflake and Google BigQuery (any SQL with bound parameters), Databricks, and Redshift through PostgreSQL.
• Graph: Neo4j (Cypher with $name parameters).
• Cloud storage and databases: AWS DynamoDB, S3-compatible object storage, Azure Blob.

Every query step has Test this step to preview rows. Large result sets stream through the flow in bounded pieces rather than being loaded at once.