Connector SDK FAQ
Do we need experienced Python developers?
No. Connector SDK connectors are written in Python, but the structure is straightforward. You can use AI coding agent plugins to scaffold the connector, run local tests, and flag issues, reducing manual writing of code. The Community Connectors Catalog also provides 100+ ready-to-use connector templates for Connector SDK.
How long does it take to build a connector?
Most teams reach a working prototype faster than expected. Once you have source credentials and API documentation available, AI coding agent plugins can scaffold the connector, run a local test, and flag issues automatically. The Community Connectors Catalog of 100+ connectors means most projects don't start from scratch.
What does Fivetran manage when I use Connector SDK?
You write and maintain the connector logic. Fivetran manages:
- Hosting and execution of your connector code
- Sync scheduling and retries
- Schema drift detection
- Monitoring and alerting
- Data type inference
- Destination table management
- Platform uptime and infrastructure
What does Fivetran handle that a custom-built pipeline wouldn't?
When you build and host a pipeline outside of Fivetran, your team is responsible for all operational concerns: scheduling, retries on failure, schema change handling, monitoring, alerting, and infrastructure maintenance. Connector SDK offloads all of that to Fivetran. Your team owns the connector logic; Fivetran owns the operations.
Can I use an LLM to build and run my own connector instead?
LLMs can generate working connector code quickly. The harder part is operating a connector reliably in production: scheduling syncs, retrying on failure, detecting and handling schema changes, monitoring for errors, and maintaining the infrastructure that runs it. When you build and host your own connector, your team owns all of that.
Connector SDK lets you use an LLM or AI coding agent plugin to write the connector code, while Fivetran handles the operational layer.
Who is responsible if a connector breaks?
Connector SDK uses a shared responsibility model. Fivetran owns the managed execution environment, scheduling, retries, schema drift detection, monitoring, alerting, logging, and infrastructure. You own the connector code: the source API logic, state handling, and fixes when the source changes its API. If you want Fivetran to manage ongoing connector health, you can buy the optional Connector SDK Maintenance add-on, which covers monitoring and updates when the source introduces breaking changes.
How secure is running custom code on Fivetran infrastructure?
Your connector code runs in a managed, isolated runtime on Fivetran infrastructure. Credentials are never stored in the connector code. They are passed at runtime through encrypted configuration values. For sources that require private network access, Fivetran supports connectivity options including PrivateLink. If you use Hybrid Deployment, your code runs in your own environment and the security model differs from the standard hosted runtime.
Who owns the connector code I build with the Connector SDK?
You write and own the connector code. Fivetran provides the runtime environment and infrastructure to execute it, but the code itself is your intellectual property.
Where is my custom connector hosted?
By default, Fivetran hosts and runs your connector in Google Cloud Platform (GCP) within the region you have configured for your destination. If you are on an Enterprise or Business Critical plan, you can choose a different cloud provider and region.
If you prefer to run the connector in your own network, you can use the Hybrid Deployment model.
For full details, see our Connector SDK architecture documentation.
What happens to my connector code and credentials when I delete a connection?
When you delete a connection, Fivetran deletes the associated connector code and all credentials stored for that connection.
Is Connector SDK the right tool for my use case?
| Situation | Recommended tool |
|---|---|
| Your source is in the Fivetran connector catalog and requires no custom logic | Native connector |
| Your source has a publicly documented REST/JSON API and you want a no-code solution | AI Connector Agent |
| Your source uses complex authentication, GraphQL, SOAP, SFTP, or a private API | Connector SDK |
| The AI Connector Agent couldn't handle your source | Connector SDK |
| You need full control over extraction logic, schema, or state management | Connector SDK |
For a full comparison of Connector SDK and the AI Connector Agent, see Connector SDK AI plugins vs. AI Connector Agent.
The AI Connector Agent didn't work for our source. What should we do?
The AI Connector Agent only supports REST APIs with publicly available documentation and doesn't handle GraphQL, SOAP, or private APIs. If your source falls into any of these categories, Connector SDK is the right tool. You can use a Connector SDK AI plugin with your preferred coding agent to accelerate development.
See Connector SDK AI plugins vs. AI Connector Agent for more details on when to use each.
What happens to our connector if the engineer who built it leaves the team?
Connector SDK connectors are standard Python files stored in your version control system. Any engineer with Python experience can read, modify, and maintain the code. Fivetran continues to run the connector on schedule regardless of team changes. The operational concerns, including scheduling, retries, monitoring, and alerting, remain with Fivetran and are not tied to any individual on your team.
Do I need to sign a partnership agreement to use the Connector SDK?
No. The Connector SDK requires no partnership agreement. You can start building immediately after creating a Fivetran account.