Fivetran MCP Server

Name: Fivetran MCP Server
Author: Fivetran

Query and manage a Fivetran account's connections, destinations, syncs and users over its REST API

Official · Fivetran Python Local · stdio Data and analytics Cloud platforms

GitHub Website

25 popularity score ★ 26

Connect your client

Local (from source)

Add in one click

Add to Cursor Add to VS Code Add to LM Studio Add to Goose

Opens the app, which asks before it adds the server. Then replace each <NAME> in its settings with your value.

Or add it by command or config

It needs FIVETRAN_API_KEY, FIVETRAN_API_SECRET. Replace each <NAME> with your value.

Claude CodeTerminal

claude mcp add fivetran -e FIVETRAN_API_KEY=<FIVETRAN_API_KEY> -e FIVETRAN_API_SECRET=<FIVETRAN_API_SECRET> -- uvx --from git+https://github.com/fivetran/fivetran-mcp fivetran-mcp

Claude Desktopclaude_desktop_config.json

{
  "mcpServers": {
    "fivetran": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/fivetran/fivetran-mcp",
        "fivetran-mcp"
      ],
      "env": {
        "FIVETRAN_API_KEY": "<FIVETRAN_API_KEY>",
        "FIVETRAN_API_SECRET": "<FIVETRAN_API_SECRET>"
      }
    }
  }
}

Cursor~/.cursor/mcp.json

{
  "mcpServers": {
    "fivetran": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/fivetran/fivetran-mcp",
        "fivetran-mcp"
      ],
      "env": {
        "FIVETRAN_API_KEY": "<FIVETRAN_API_KEY>",
        "FIVETRAN_API_SECRET": "<FIVETRAN_API_SECRET>"
      }
    }
  }
}

VS Code.vscode/mcp.json

{
  "servers": {
    "fivetran": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/fivetran/fivetran-mcp",
        "fivetran-mcp"
      ],
      "env": {
        "FIVETRAN_API_KEY": "<FIVETRAN_API_KEY>",
        "FIVETRAN_API_SECRET": "<FIVETRAN_API_SECRET>"
      }
    }
  }
}

Codex~/.codex/config.toml

[mcp_servers.fivetran]
command = "uvx"
args = ["--from", "git+https://github.com/fivetran/fivetran-mcp", "fivetran-mcp"]

[mcp_servers.fivetran.env]
FIVETRAN_API_KEY = "<FIVETRAN_API_KEY>"
FIVETRAN_API_SECRET = "<FIVETRAN_API_SECRET>"

Gemini CLITerminal

gemini mcp add --scope user -e FIVETRAN_API_KEY=<FIVETRAN_API_KEY> -e FIVETRAN_API_SECRET=<FIVETRAN_API_SECRET> fivetran uvx -- --from git+https://github.com/fivetran/fivetran-mcp fivetran-mcp

Settings 10

`FIVETRAN_API_KEY`	Required: Fivetran API key from the dashboard's API config page (stdio mode only).
`FIVETRAN_API_SECRET`	Required: Matching Fivetran API secret (stdio mode only).
`FIVETRAN_SCOPE`	Optional: read (default), read/write or read/write/delete; the ceiling on what tools are exposed.
`DISALLOWED_ACTIONS`	Optional: Comma-separated resource:action or resource:action:endpoint tokens to deny, e.g. the recommended credential-endpoint denylist.
`FIVETRAN_ALLOW_WRITES`	Optional: Older flag; true equals FIVETRAN_SCOPE=read/write. Ignored if FIVETRAN_SCOPE is set.
`MCP_TRANSPORT`	Optional: stdio (default) or streamable-http.
`MCP_HOST`	Optional: Bind host for streamable-http mode; default 127.0.0.1.
`MCP_PORT`	Optional: Bind port for streamable-http mode; default 8000.
`MCP_ALLOWED_ORIGINS`	Optional: Allowed Origin header values in HTTP mode; without it or MCP_ALLOWED_HOSTS there is no DNS-rebinding protection.
`MCP_ALLOWED_HOSTS`	Optional: Allowed Host header values in HTTP mode.

What it does

Fivetran MCP Server is Fivetran's own bridge between an AI client and the Fivetran REST API. Out of the box it answers questions about an account: which connections are failing, when a given connection last synced, which destinations and groups exist.

Rather than one hand-written tool per task, it carries a manifest of the whole Fivetran API. Two discovery tools let the client find an endpoint and read its schema (including the service-specific config fields needed to create, say, a Postgres connection), and one execution tool per resource and action pair, such as connections_read or destinations_delete, calls it. The execution tools a client sees depend on FIVETRAN_SCOPE (read by default; read/write or read/write/delete to go further) and on a DISALLOWED_ACTIONS denylist that can remove whole resources or single endpoints.

It normally runs as a local stdio process fetched with uvx from the GitHub repository. Since version 0.3 it can also run as a streamable HTTP service that several clients share, with each request carrying its own Fivetran credentials. The makers note that OAuth resource-server mode is scaffolding only and does not work yet. Fivetran also publishes plugins for Claude Code and Codex built on this server, such as copy-connections for copying connections to a new destination.

What it needs

Runtime	Python 3.10+

Python 3.10 or newer and uv (for uvx), or a local clone installed with pip. A Fivetran account and an API key and secret generated in the Fivetran dashboard's API config page; what it can do is bounded by that user's permissions in Fivetran.

Limits

Read-only unless FIVETRAN_SCOPE is raised. Write and delete tools carry advisory notes asking the client to confirm with the user, but the server does not enforce confirmation. It is not on PyPI: the fivetran-mcp and mcp-fivetran names there belong to unrelated projects, so plain uvx fivetran-mcp does not work. OAuth in HTTP mode is not functional yet; HTTP mode only forwards each request's Basic auth header, and refuses to start if FIVETRAN_API_KEY or FIVETRAN_API_SECRET are set.

Who it suits

Teams already on Fivetran who want to check sync health or troubleshoot connections from a chat client, and, with a raised scope, to create, pause, resync or copy connections without the dashboard. Because it is official and exposes the full API, it suits admins who will set the scope and denylist with care; keep the default read scope for anyone who only needs status. It manages pipelines only; it does not query the data warehouse those pipelines load, so pair it with a server for your destination database for that.

Tools 53

list_endpoints: Finds Fivetran API endpoints by category or search term, or counts them per category.
get_schema: Shows an endpoint's parameters, request body and response shape, with per-service config fields.
account_read: Runs read endpoints for account details (1 endpoint).
connections_delete: Runs delete endpoints for connections (5 endpoints).
connections_read: Runs read endpoints for connections (12 endpoints).
connections_write: Runs create, change or sync endpoints for connections (18 endpoints).
connector_sdk_delete: Runs delete endpoints for Connector SDK packages (1 endpoint).
connector_sdk_read: Runs read endpoints for Connector SDK packages (3 endpoints).
connector_sdk_write: Runs create, change or sync endpoints for Connector SDK packages (2 endpoints).
destinations_delete: Runs delete endpoints for destinations (3 endpoints).
destinations_read: Runs read endpoints for destinations (6 endpoints).
destinations_write: Runs create, change or sync endpoints for destinations (5 endpoints).
external_logging_delete: Runs delete endpoints for external logging services (2 endpoints).
external_logging_read: Runs read endpoints for external logging services (3 endpoints).
external_logging_write: Runs create, change or sync endpoints for external logging services (6 endpoints).
external_secrets_managers_delete: Runs delete endpoints for external secrets managers (1 endpoint).
external_secrets_managers_read: Runs read endpoints for external secrets managers (3 endpoints).
external_secrets_managers_write: Runs create, change or sync endpoints for external secrets managers (2 endpoints).
external_secrets_managers_entities_read: Runs read endpoints for external secrets manager entities (1 endpoint).
groups_delete: Runs delete endpoints for groups (2 endpoints).
groups_read: Runs read endpoints for groups (6 endpoints).
groups_write: Runs create, change or sync endpoints for groups (3 endpoints).
hvr_write: Runs create, change or sync endpoints for HVR (1 endpoint).
hybrid_deployment_agents_delete: Runs delete endpoints for hybrid deployment agents (1 endpoint).
hybrid_deployment_agents_read: Runs read endpoints for hybrid deployment agents (2 endpoints).
hybrid_deployment_agents_write: Runs create, change or sync endpoints for hybrid deployment agents (3 endpoints).
metadata_read: Runs read endpoints for connector metadata (2 endpoints).
private_links_delete: Runs delete endpoints for private links (1 endpoint).
private_links_read: Runs read endpoints for private links (2 endpoints).
private_links_write: Runs create, change or sync endpoints for private links (2 endpoints).
proxy_delete: Runs delete endpoints for proxy agents (1 endpoint).
proxy_read: Runs read endpoints for proxy agents (3 endpoints).
proxy_write: Runs create, change or sync endpoints for proxy agents (2 endpoints).
public_read: Runs read endpoints for public connector type (1 endpoint).
roles_read: Runs read endpoints for roles (1 endpoint).
system_keys_delete: Runs delete endpoints for system keys (1 endpoint).
system_keys_read: Runs read endpoints for system keys (2 endpoints).
system_keys_write: Runs create, change or sync endpoints for system keys (3 endpoints).
teams_delete: Runs delete endpoints for teams (5 endpoints).
teams_read: Runs read endpoints for teams (8 endpoints).
teams_write: Runs create, change or sync endpoints for teams (8 endpoints).
transformation_projects_delete: Runs delete endpoints for transformation projects (1 endpoint).
transformation_projects_read: Runs read endpoints for transformation projects (2 endpoints).
transformation_projects_write: Runs create, change or sync endpoints for transformation projects (3 endpoints).
transformations_delete: Runs delete endpoints for transformations (1 endpoint).
transformations_read: Runs read endpoints for transformations (4 endpoints).
transformations_write: Runs create, change or sync endpoints for transformations (5 endpoints).
users_delete: Runs delete endpoints for users (5 endpoints).
users_read: Runs read endpoints for users (8 endpoints).
users_write: Runs create, change or sync endpoints for users (8 endpoints).
webhooks_delete: Runs delete endpoints for webhooks (1 endpoint).
webhooks_read: Runs read endpoints for webhooks (2 endpoints).
webhooks_write: Runs create, change or sync endpoints for webhooks (4 endpoints).

Popularity

25 popularity score

★ 26

GitHub stars	26	17.2 points
Official server	Yes	8 points

Measured 2026-11-07. How the score works.

Where it is published

Code	fivetran/fivetran-mcp ★ 26

What it can reach

Runs	On your machine
Reaches	The internet, Your account at a service, Cloud resources
Changes things	Yes: it can write, send, create or delete
Sign-in	An API key or token
Licence	MIT (GitHub)
Latest version	0.5.1 GitHub release, 2026-10-06
Last code change	2026-11-04

Runs on your machine (or your own host in HTTP mode) and calls the Fivetran REST API with the API key and secret you give it, acting with that Fivetran user's permissions. Default scope is read; read/write and read/write/delete expose tools that create, modify, sync and delete connections, destinations, users, teams, webhooks and more. The manifest covers every API endpoint, including ones that read, create or rotate system keys and user API keys that outlive the session; the makers recommend denying these with DISALLOWED_ACTIONS and the server warns at startup while they are callable. Confirmation before writes is only an advisory note to the client. In HTTP mode, binding beyond localhost without MCP_ALLOWED_ORIGINS or MCP_ALLOWED_HOSTS leaves it without DNS-rebinding protection.

Over time

GitHub stars 26 +0

2626

14 Oct 20267 Nov 2026

Popularity score 25 −0

2525

14 Oct 20267 Nov 2026

What is rising across the directory

Sources

Page checked 2026-10-14.

Badge for your README

Shows this server's rank in its category, and stars for a server that runs on your machine. It updates every hour.

MarkdownREADME.md

[![Fivetran MCP Server on EveryMCPServer](https://everymcpserver.com/badge/fivetran-mcp-server.svg)](https://everymcpserver.com/servers/fivetran-mcp-server)

Connect your client

Add in one click

Or add it by command or config

Settings 10

What it does

What it needs

Limits

Who it suits

Tools 53

Similar servers

StackQL MCP Server

PostHog MCP Server

Harness MCP Server

MCP DigitalOcean Integration

Databuddy MCP server

Metabase MCP server