Fivetran Connector Sdk

Fivetran Connector Builder — Claude Code Plugin

Build, test, fix, and deploy Fivetran connectors with AI assistance, directly in Claude Code.

Prerequisites

• Python 3.10-3.14
• Claude Code installed
• Fivetran Connector SDK — pip install fivetran-connector-sdk
• A Fivetran account (https://fivetran.com)

Installation

See the top-level README for the full install matrix. Quick path:

claude plugin marketplace add fivetran/connector_sdk_tools
claude plugin install fivetran-connector-sdk@fivetran-connector-sdk-ai

Or from inside a Claude Code session:

/plugin marketplace add fivetran/connector_sdk_tools
/plugin install fivetran-connector-sdk@fivetran-connector-sdk-ai

Post-installation

Install the tool dependencies:

macOS/Linux:

pip install -r /path/to/claude-code/tools/requirements.txt

Windows PowerShell:

python -m pip install -r "C:\path\to\claude-code\tools\requirements.txt"

Quick Start Tutorial

Step 1: Start Claude Code with the Plugin

cd ~/my-connectors  # or wherever you want to create connectors
claude --plugin-dir /path/to/claude-code

Step 2: Build Your First Connector

In the Claude Code chat, describe the connector you want to build:

/build-connector <describe your data source and what tables you want to sync>

Claude will:

1. Research the API documentation
2. Ask you clarifying questions about scope and data requirements
3. Generate the connector code
4. Set up a Python virtual environment
5. Ask you to enter your API credentials

Step 3: Enter Your API Credentials

When prompted, open a separate terminal and run:

macOS/Linux:

cd ~/my-connectors/your_connector
python -m pip install -r /path/to/claude-code/tools/requirements.txt
python /path/to/claude-code/tools/enter_configuration.py configuration.json

Windows PowerShell:

cd "C:\path\to\my-connectors\your_connector"
python -m pip install -r "C:\path\to\claude-code\tools\requirements.txt"
python "C:\path\to\claude-code\tools\enter_configuration.py" "configuration.json"

Enter configuration values when prompted. Every field value is encrypted immediately as an inline ENCRYPTED:v1:<key_id>:local-fernet: value in configuration.json. Claude never sees plaintext configuration values.

The dependency install is temporary. Once secure configuration entry is available in the Fivetran Connector SDK CLI, this helper script flow will be replaced.

First time only: The tool creates a local encryption secret under your user profile (~/.fivetran/csdk_master_secret on macOS/Linux, %USERPROFILE%\.fivetran\csdk_master_secret on Windows) and uses it immediately. Claude never sees the secret or plaintext configuration values.

Only enter_configuration.py creates the secret. The test and deploy tools require the existing secret to decrypt configuration values at runtime. To start configuration entry over, run enter_configuration.py again.

Go back to Claude Code and tell it you've entered your credentials.

Step 4: Test the Connector

Claude will automatically run the first test. If it succeeds, you'll see:

• Tables synced
• Row counts for each table
• Sample data

If it fails, Claude will diagnose the issue:

• INFRA error: Network/infrastructure problem
• FIRST_RUN error: Likely a credentials or config issue — verify your API credentials
• CODE error: Claude will automatically fix the code and re-test

Step 5: Iterate and Deploy

Once testing passes, you can:

• Ask Claude to add more tables or features
• Run /test-connector anytime to re-test
• Run /migrate-functions-connector to port an existing Fivetran Functions connector to Connector SDK
• Run /migrate-meltano-connector to port an existing Meltano extractor or Singer tap to Connector SDK
• Run /migrate-airbyte-connector to port an existing Airbyte source connector to Connector SDK
• Run /deploy-connector when ready to deploy to Fivetran

Example Session

You: /build-connector <Your API> connector for <tables you want>

Claude: I'll build a connector for <Your API>. Let me research the API...
        [researches API docs]

        I have a few questions:
        1. <clarifying question about scope>
        2. <clarifying question about filtering>

You: <your answers>

Claude: Got it. Let me generate the connector...
        [creates <your_connector>/ directory with connector.py, configuration.json, README.md]
        [sets up virtual environment]

        Please enter your API credentials. In a separate terminal, run:
          python .../enter_configuration.py configuration.json

You: Done, I've entered my credentials

Claude: Running the connector test...
        [runs fivetran debug]

        ✅ Test passed!
        Tables synced:
          - <table_1>: X rows
          - <table_2>: Y rows

        Your connector is ready. Would you like to deploy it to Fivetran?

Usage

Build a New Connector

/build-connector Stripe API connector for payments, customers, and invoices

This will:

1. Research the API documentation
2. Ask clarifying questions if needed
3. Generate connector code following SDK best practices
4. Set up a virtual environment
5. Run an initial test
6. Auto-fix any code errors
7. Prompt you to enter configuration values (API credentials)

Entering Configuration Values

After the connector is generated, you'll need to enter your API credentials. Run this in a separate terminal:

python -m pip install -r /path/to/claude-code/tools/requirements.txt
python /path/to/claude-code/tools/enter_configuration.py configuration.json

On first run, it creates a local encryption secret under your user profile. This encrypts local configuration values so the AI cannot see them.

Test a Connector

/test-connector

Runs fivetran debug, checks warehouse.db for synced data, and reports results.

Fix or Modify a Connector

Just describe the problem or change in natural language — no slash command needed:

I'm getting an authentication error: Invalid API key

or

Add a "pull requests" table to my GitHub connector

The plugin routes to the connector-fixer subagent, which classifies the error (INFRA / FIRST_RUN / CODE), researches the correct SDK pattern, and applies targeted fixes to connector.py.

Deploy a Connector

/deploy-connector

Validates, runs a final test, and guides you through Fivetran deployment.

What's Included

Component	Description
/build-connector	Full generation workflow (research → generate → test → auto-fix)
/test-connector	Run and validate connector tests
/deploy-connector	Package and deploy to Fivetran
/migrate-functions-connector	Migrate Fivetran Functions connectors to Connector SDK
/migrate-meltano-connector	Migrate Meltano extractors or Singer taps to Connector SDK
/migrate-airbyte-connector	Migrate Airbyte source connectors to Connector SDK
agents/connector-validator.md	Subagent for API research and requirements gathering
agents/connector-generator.md	Subagent for generating connector code
agents/connector-fixer.md	Subagent for diagnosing and fixing errors (invoked automatically on natural-language fix requests)
tools/enter_configuration.py	Enter and encrypt API credentials
tools/run_connector.py	Run connector with runtime config via named pipe
tools/deploy_connector.py	Deploy connector with runtime config via named pipe

How It Works

This plugin turns Claude Code into a Fivetran connector development tool by injecting:

1. Domain knowledge — Fivetran SDK patterns, best practices, and common pitfalls
2. Workflow skills — Step-by-step processes for building, testing, fixing, and deploying
3. Specialized agents — Subagents focused on research, generation, and debugging
4. SDK examples — Automatic fetching of relevant patterns from the official examples repo

Claude Code provides the AI conversation, file editing, terminal access, and git integration. This plugin adds the Fivetran-specific expertise on top.

Telemetry

This plugin collects anonymous usage data (skill name, plugin name and version, model, status (started, ok, or fail), session ID, timestamp) to help improve the product. No prompts, code, or personal information are collected. To opt out, add to your shell profile:

export FIVETRAN_TELEMETRY_DISABLED=1