How to connect Fivetran to any data source

One of the most common questions data teams ask when evaluating Fivetran is: what happens when my source isn't in the connector catalog? It's a fair question. With over 750 connectors, Fivetran covers an enormous range of sources, but every organization has at least one system that doesn't appear on a standard connector list. A proprietary CRM built in-house. An ERP with no public API. A partner data feed arriving as EDI files. A legacy application that only exposes its database layer.

The answer is that Fivetran isn't just a connector catalog; it's a connectivity platform with 5 distinct paths for getting data from any source into your destination. Each path suits a different type of source and a different level of engineering investment. This post maps each clearly so you can pick the right approach for every source in your estate, not just the ones that appear in the directory.

[CTA_MODULE]

 1. Managed Connector

A Managed Connector is Fivetran's fully engineered, continuously maintained integration for a specific source. With over 750 connectors spanning SaaS applications, ERPs, marketing platforms, advertising networks, CRMs, HR systems, and event sources, this is the right starting point for the overwhelming majority of modern business data sources. Standard connectors — built for well-established platforms like Salesforce, HubSpot, Shopify, Google Ads, NetSuite, Workday, and Snowflake — go deep into each platform's API, surfacing all available objects and fields. Lite connectors extend coverage to niche or newer SaaS applications at an accelerated build pace. Partner-built connectors are maintained by technology partners for their own platforms.

The managed model means Fivetran's engineering team owns every connector's ongoing maintenance. When Salesforce releases API v65.0, Fivetran upgrades the connector — not your team. When a marketing platform changes its pagination strategy, Fivetran handles it. Connection is typically via OAuth, requiring no API knowledge from the user. Fivetran applies a sync-all strategy: it discovers and replicates all accessible objects automatically, adapts to schema changes as they happen, and handles incremental sync using the source's native change tracking mechanisms.

2. File Connector

A large category of enterprise data never goes through an API. Financial exports, vendor data feeds, EDI transaction files, HL7 clinical messages, regulatory submissions, legacy system extracts — these are all delivered as files. They land in S3 buckets, SFTP servers, Google Drive folders, Azure Blob Storage, or on internal network shares. File Connectors are Fivetran's answer to this category, and they handle significantly more complexity than the label suggests.

Fivetran's File Connectors support every major structured format — CSV with configurable delimiters, JSON Lines, Parquet, and XML — plus major compression formats including gzip, zip, bz2, and Zstandard. Automatic schema inference means Fivetran reads the file structure without requiring upfront mapping. Metadata columns (sync timestamp, source file path, row identifier) are added to every row. User-defined primary keys handle deduplication. Automatic type detection and schema evolution manage changing file structures over time. For files on internal networks or on-premises SFTP servers, Fivetran's Hybrid Deployment agent reads files from within the network boundary — no cloud exposure of the raw data required.

3. Database Connector

Many applications don't expose a clean API, or their API lacks the depth or historical coverage that analytics requires. But almost every application runs on a database, and that database is often the richest, most complete representation of the application's data, including historical records, deleted rows, and fields that the API never surfaces. Fivetran's Database Connectors connect directly to the underlying database using log-based change data capture (CDC), reading committed changes directly from the transaction log rather than querying the application layer.

Fivetran supports the major database engines in this model: PostgreSQL MySQL and MariaDB (binary log), SQL Server (CDC tables), Oracle (binary log reader), MongoDB, DB2, and more. Binary Log-based CDC is the preferred approach — it captures every insert, update, and delete with minimal impact on the source database because it reads asynchronously from the transaction log rather than running SELECT queries against production tables. Deploying using Fivetran Hybrid ensures all network connections are outbound from the host network; this is the safest way to deploy.

4. Connector SDK

The Connector SDK is the path for sources that have no Managed Connector, can't be reached via a database, and don't deliver files. Internal APIs, proprietary vendor platforms, niche industry applications, bespoke authentication schemes, and legacy systems with unusual access patterns all fall into this category. Before the SDK existed, connecting these sources meant writing and hosting entirely custom ETL scripts — with all the operational overhead that implies: scheduling, monitoring, error handling, schema management, state persistence, retry logic, and infrastructure maintenance. The SDK eliminates all of that.

SDK connectors are written in Python using a structured framework that handles the operational plumbing automatically. You implement 2 functions: a schema function (defines the tables and columns to be created in the destination) and an update function (contains the extraction logic — call the API, yield records, save a cursor for the next incremental sync). Fivetran runs the connector on its own managed infrastructure, handling scheduling, error recovery, checkpointing, schema evolution, and dashboard monitoring. Most connectors can be built and deployed in a few hours. Completed SDK connectors appear in the Fivetran dashboard alongside every built-in connector — same observability, same schema management, same reliability guarantees. The SDK has also become the mechanism by which technology partners publish new connectors to the Fivetran ecosystem.

5. Webhook Connector

Webhooks invert the data collection model. Instead of Fivetran pulling data from a source on a schedule, the source pushes data to Fivetran the moment something happens. For event-driven systems, like payment processors firing on each transaction, e-commerce platforms pushing order events, monitoring tools alerting on incidents, IoT devices sending sensor readings, and CRMs notifying on deal stage changes, this push model captures every event in real time, without polling lag and without requiring Fivetran to maintain an ongoing connection to the source system.

Fivetran's Webhook Connector provides a unique HTTPS endpoint for each connection. Any system or service that can make an HTTP POST request, which is virtually every modern platform, can send data to this endpoint in JSON or as query parameters. Fivetran ingests the payload, infers the schema, and loads it into the destination. There's no SDK to write, no authentication flow to configure, no API to poll. If the source already has webhook functionality (most SaaS platforms do — Stripe, GitHub, Shopify, PagerDuty, Segment, and hundreds of others all support outbound webhooks), this is often the fastest possible path to getting event data into the warehouse. The Webhook Connector is also the universal bridge for any tool that doesn't fit the other 4 paths. So if it can talk HTTP, it can talk to Fivetran.

No source is unreachable: Pick the right path 

Managed Connectors for the broad catalog of SaaS and ERP platforms. File Connectors for data that arrives as files. Database Connectors for direct access to application databases via log-based CDC. The Connector SDK for proprietary APIs and internal systems. Webhooks for real-time event streams and anything that can talk HTTP. Together, these 5 paths mean there is no data source that Fivetran cannot connect to — only the question of which method requires the least engineering effort for your specific situation.

[CTA_MODULE]