Integrations
Connect the source once. Map every file that lands.
A connector is a scheduled read of somewhere your data already sits. Each run pulls what changed, runs the cascade against your schema, and writes the result where you point it — with the credential held encrypted and never echoed back.
- Secrets envelope-encrypted with a KEK
- Never accepted in a request body
- Incremental watermarks, not full re-reads
- Dry-run before anything is written
- Region pinned per workspace
Object storage
Object storage sources
Amazon S3
SigV4 read + write against S3 and any S3-compatible endpoint. Incremental by object modified-time.
s3
Read the reference →Google Cloud Storage
Service-account auth over the GCS JSON API. Read the newest object under a prefix, or land the result as a file.
gcs
Read the reference →Azure Blob Storage
SharedKey auth against a storage account. Read the newest blob under a prefix, or land the result as a file.
azure_blob
Read the reference →Databases
Databases sources
SQL over HTTP
Read with your own SQL, write with generated INSERTs, over any HTTPS query endpoint. Incremental via a {{since}} token.
sql_http · sql_write
Read the reference →Supabase
Batched row inserts into PostgREST. Upsert on named columns, return=minimal, errors surfaced verbatim.
supabase
Read the reference →Warehouses
Warehouses sources
BigQuery
Service-account streaming insert into a dataset table. Per-row rejections surfaced, never counted as written.
bigquery
Read the reference →Snowflake
SQL Statement API v2 with key-pair JWT auth. Batched inserts, independent batches, honest partial failures.
snowflake
Read the reference →Databricks
SQL Statement Execution API against a named warehouse. Bearer token auth, batched inserts, named failures.
databricks
Read the reference →Not listed
Anything that can produce a file already works.
A connector is a convenience, not a requirement. Every capability on this site is reachable by posting the file itself — the integrations above exist so a recurring drop does not need a human in the loop.
Ready when you are
One key, every source, one schema out.
Start with a $10 prepaid wallet. Every map draws a few tokens — including a fully deterministic one.