CDK overview
The CDK provides:- Spec contracts – define capabilities (batch extract, load, streaming) and configuration schema
- Runtime hooks – implement list, read, write, and test methods
- Packaging – bundle dependencies within supported constraints
- Local testing – run against sample payloads before publishing to your workspace
Custom connectors inherit your organization’s security review: secrets never belong in source code; use connection parameters and vault references.
Python CDK
Author connectors in Python when you need rapid iteration and rich HTTP/SDK libraries:Scaffold the project
Use the CDK template to generate package layout,
connector.yaml, and sample tests.Discovery
Return object and field metadata dynamically or from static definitions when APIs are thin.
Read/write paths
Implement pagination, backoff, and checkpointing for incremental extracts; respect destination batch limits on load.
Validation
Before publishing, run validation suites that check:- Schema of configuration JSON matches the declared form
- Required methods exist for advertised capabilities
- Dependency versions stay within allowlisted ranges
Publishing
Publish connectors to your private registry or workspace library depending on deployment. Tag versions semantically (1.4.2); pipelines pin connector versions so upgrades are explicit.
- Staging rollout
- Production promotion
Install the connector in a staging workspace first; run synthetic and mirrored production workloads.
Related topics
Connector gallery
Compare first-party coverage before building custom.
API keys
Secure programmatic testing of connectors.