Skip to content
Fivetran MCP Server

Fivetran MCP Server

by Fivetran
Released
MIT
Repository Docs
Databases & Data

Summary

Fivetran's own MCP server exposes the Fivetran REST API to agents as discoverable tools, so an assistant can check sync status, inspect connections and destinations, and trigger syncs from chat.

Features

  • Discovery-first tool design: list_endpoints and get_schema reach the whole Fivetran REST API without flooding the client's tool list
  • Read-only by default; write and delete are enabled explicitly through FIVETRAN_SCOPE
  • DISALLOWED_ACTIONS denylist blocks individual operations, including the API-key endpoints Fivetran recommends denying
  • Runs with uvx straight from the repository, or as a hosted streamable-HTTP service for a team
  • Connector-specific configuration schemas via get_schema with a service parameter

Installation

Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.

Any MCP-compatible agent

Most agents (Claude, Cursor, Windsurf, VS Code, and more) read a standard mcpServers configuration.

  1. Open your agent's MCP configuration file.
  2. Merge the snippet below into it, filling in the environment variables with your own values.
  3. Restart the agent — the "Fivetran MCP Server" tools will be available.
{
  "mcpServers": {
    "fivetran-mcp": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/fivetran/fivetran-mcp",
        "fivetran-mcp"
      ],
      "env": {
        "FIVETRAN_API_KEY": "your-api-key",
        "FIVETRAN_API_SECRET": "your-api-secret",
        "FIVETRAN_SCOPE": "read",
        "DISALLOWED_ACTIONS": "system-keys:read,users:read:get_user_api_key,users:read:list_api_keys,users:write:create_user_api_key,users:write:rotate_user_api_key,users:delete:delete_user_api_keys"
      }
    }
  }
}

Description

Fivetran moves data from hundreds of sources into a warehouse, and most of the day-to-day questions about it are small and awkward: did the Postgres connector finish, which connections are broken, when did this destination last load. Answering them means opening the dashboard and clicking through connector pages. The Fivetran MCP Server is Fivetran's first-party Model Context Protocol server that puts those questions inside whatever MCP client you already work in.

How it exposes the API

Rather than registering one tool per endpoint — which would flood a client's tool list — the server uses a discovery-then-execute shape:

  • list_endpoints browses the available API operations by category or free-text search.
  • get_schema returns the parameter and response schema for a chosen operation, and with a service parameter returns the connector-specific configuration fields for a particular source.
  • Execution tools are resource/action pairs such as connections_read, connections_write and destinations_delete.

That design keeps the tool surface small while still reaching the whole API, and it means the agent reads the real schema before it constructs a call.

Permissions are opt-in, not implied

The server starts read-only. FIVETRAN_SCOPE widens it to read/write or read/write/delete, and DISALLOWED_ACTIONS denies individual operations by resource:action. The README's recommended denylist blocks the credential-adjacent endpoints — system-keys:read, the user API-key read/create/rotate/delete operations — so an agent with write scope still cannot mint itself a key. Setting that denylist is worth doing before the first run rather than after.

Running it

It is a Python package (3.10+) run through uv, with no install step required:

uvx --from git+https://github.com/fivetran/fivetran-mcp fivetran-mcp

It authenticates with a Fivetran API key and secret. For team use it can also run as a hosted HTTP service with --transport streamable-http --host 0.0.0.0 --port 8000, and the repository ships plugin repositories (for example copy-connections) that add higher-level operations on top.

Who it is for

Data and analytics engineers who already run Fivetran and want incident triage and routine connector questions answered without a context switch. It is not a replacement for the Fivetran UI — it is the same API, addressed in language, with a permission model you configure up front.

MIT licensed, maintained in Fivetran's own GitHub organisation.

Related MCP Servers

See all Fivetran MCP Server alternatives

MCP: Astro Airflow MCP Server

by Astronomer

Astronomer's MCP server for Apache Airflow: 30+ tools over the Airflow REST API, with consolidated helpers that explore a DAG, diagnose a failed run or summarise system health in a single call. Works with Airflow 2.x and 3.x.

Cloud & Infrastructure

MCP: Keboola MCP Server

by Keboola

Opens a Keboola project to an agent: query storage tables, write SQL transformations, configure extractors and writers, trigger jobs and build flows.

Databases & Data

MCP: Couchbase Guru

by Couchbase

Couchbase's documentation agent as a single MCP tool: ask a question about any Couchbase product, SDK or version and get an answer with source links. No cluster and no credentials.

Databases & Data
Browse all MCP servers →