Fivetran MCP Server

A read-only MCP (Model Context Protocol) server for troubleshooting Fivetran connectors. Connect it to Cursor (or any MCP client) and diagnose sync issues, check connector health, and inspect configurations using natural language.

What It Does

• Lists and filters connectors by environment and health status
• Diagnoses connector issues with actionable recommendations
• Shows environment-level health summaries at a glance
• Inspects connector configuration (with sensitive fields redacted)
• Monitors hybrid deployment agents

Read-only — no modifications to your Fivetran setup, ever.

Available Tools

Tool	What it does
list_connectors	List connectors, filter by env and status
get_group_health_summary	Dashboard-style overview of an environment
diagnose_connector	Health check with severity-ranked issues and recommendations
get_connector_config	Full connector configuration (credentials redacted)
get_connector_schema_status	Table-level sync status
get_sync_history	Sync timestamps, warnings, and error details
list_hybrid_agents	List hybrid deployment agents
get_hybrid_agent_details	Detailed agent status and assigned connectors

Quick Start

Prerequisites

• Python 3.12+
• uv package manager
• Fivetran API credentials (Get them here)

1. Install

git clone https://github.com/redhat-data-and-ai/fivetran-mcp-server.git
cd fivetran-mcp-server
make install

2. Configure

cp .env.example .env

Edit .env and add your Fivetran API key and secret:

FIVETRAN_API_KEY=your-api-key
FIVETRAN_API_SECRET=your-api-secret

3. Run

make local

The server starts at http://localhost:8080. Verify with:

curl http://localhost:8080/health

4. Connect to Cursor

Add to .cursor/mcp.json in your project:

{
  "mcpServers": {
    "fivetran": {
      "url": "http://localhost:8080/mcp"
    }
  }
}

Reload Cursor (Cmd+Shift+P → "Developer: Reload Window"), then ask:

• "Show me failed connectors in prod"
• "Diagnose connector shouldn_snack"
• "Give me a health summary for dev"
• "What's the config for connector abc123?"
• "List offline hybrid agents"

Tool Reference

list_connectors

List connectors filtered by environment and/or health status.

list_connectors(env="prod", status="failed")

Parameter	Values	Description
env	"dev", "preprod", "prod", "sandbox", or group ID	Partial match on group name
status	"all", "failed", "healthy", "paused", "warning"	Filter by health

get_group_health_summary

Get an environment-level dashboard: connector counts by status plus the top 5 worst offenders.

get_group_health_summary(env="prod")

diagnose_connector

Comprehensive health check with severity-ranked issues and actionable recommendations.

diagnose_connector(connector_id="abc123")

get_connector_config

Inspect full connector configuration (networking, schedule, source settings). Sensitive fields like passwords, tokens, and keys are automatically redacted.

get_connector_config(connector_id="abc123")

get_connector_schema_status

See which tables are enabled/disabled and their sync modes.

get_connector_schema_status(connector_id="abc123")

get_sync_history

Get sync timestamps, active warnings with error messages, and optionally the sync schedule config.

get_sync_history(connector_id="abc123")
get_sync_history(connector_id="abc123", include_config=True)

list_hybrid_agents

List hybrid deployment agents filtered by environment and/or connection status.

list_hybrid_agents(env="prod", status="live")

Parameter	Values	Description
env	"dev", "preprod", "prod", "sandbox"	Partial match on group name
status	"all", "live", "offline"	Filter by connection status

get_hybrid_agent_details

Get detailed info for a specific agent including assigned connectors.

get_hybrid_agent_details(agent_id="abc123")

Docker

# Pre-built image
docker pull ghcr.io/redhat-data-and-ai/fivetran-mcp-server:latest

docker run -p 8080:8080 \
  -e FIVETRAN_API_KEY=your_key \
  -e FIVETRAN_API_SECRET=your_secret \
  ghcr.io/redhat-data-and-ai/fivetran-mcp-server:latest

Or build your own:

docker build -t fivetran-mcp-server -f Containerfile .

Configuration

Variable	Default	Description
FIVETRAN_API_KEY	(required)	Fivetran API key
FIVETRAN_API_SECRET	(required)	Fivetran API secret
FIVETRAN_BASE_URL	https://api.fivetran.com/v1	API base URL
MCP_HOST	localhost	Server bind address
MCP_PORT	8080	Server port
MCP_TRANSPORT_PROTOCOL	http	Transport: http, streamable-http, or sse
PYTHON_LOG_LEVEL	INFO	Log level

Development

make install          # Set up venv + deps + pre-commit hooks
make test             # Run tests
make local            # Start the server locally

Running Tests

.venv/bin/python -m pytest                              # All tests
.venv/bin/python -m pytest --cov=fivetran_mcp_server    # With coverage
.venv/bin/python -m pytest tests/test_connectors.py -v  # Specific file

Code Quality

ruff check .    # Lint
ruff format .   # Format

Project Structure

fivetran-mcp-server/
├── fivetran_mcp_server/
│   ├── main.py              # Entry point (uvicorn)
│   ├── api.py               # FastAPI app + health endpoint
│   ├── mcp.py               # MCP server + tool registration
│   ├── settings.py          # Pydantic settings + validation
│   ├── fivetran_client.py   # HTTP client (retries, timeouts, pooling)
│   ├── tools/
│   │   ├── __init__.py      # Shared error handling decorator
│   │   └── connectors.py    # All MCP tool implementations
│   └── utils/
│       └── pylogger.py      # Structured logging (structlog)
├── tests/                   # Pytest suite (138 tests, 80%+ coverage)
├── .env.example             # Environment template
├── Containerfile            # OCI container build
├── Makefile                 # Dev commands
└── pyproject.toml           # Dependencies + tool config

License

Apache 2.0