Creativity · MCP — server

MCP Fivetran Server

The MCP Fivetran Server integrates Fivetran's REST API so an MCP client can inventory data connectors, check last-sync timestamps, fetch sync logs, and trigger a resync for a failing connector. It shortens on-call loops for analytics engineering teams by turning dashboard tasks into conversational steps.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: Python
Transports: stdio

Capabilities

Tools: list_connectors, get_connector_status, trigger_resync, fetch_sync_logs, update_schedule
Resources: fivetran://connector/{id}
Auth: Fivetran API key + API secret (HTTP basic)

Install

pipx install mcp-server-fivetran

Configuration

{
  "mcpServers": {
    "fivetran": {
      "command": "mcp-server-fivetran",
      "env": {
        "FIVETRAN_API_KEY": "${FIVETRAN_API_KEY}",
        "FIVETRAN_API_SECRET": "${FIVETRAN_API_SECRET}"
      }
    }
  }
}

Frequently asked questions

Is Fivetran suitable for ad-hoc syncs?

Fivetran is best for scheduled incremental syncs of production systems. Triggering a one-off resync from an LLM is supported but should be an exception rather than the default workflow.

Can it create new connectors?

Yes, via create_connector, but most teams provision connectors through Terraform or the UI. Creating via MCP is more useful in demo/sandbox scenarios.

How does it compare to Airbyte MCP?

Fivetran is managed SaaS with a fixed connector catalog. Airbyte is OSS/self-hostable with a broader, more editable catalog. MCP shapes are similar.

Sources

Fivetran REST API — accessed 2026-04-20
Model Context Protocol — accessed 2026-04-20
MCP servers repo — accessed 2026-04-20