Best Open Source Connector Catalogs for Product Teams
Airbyte, Meltano Hub and dlt compared on connector count, license and how you build the long tail, with Fivetran as the managed benchmark
A connector catalog is the set of prebuilt sources and destinations an ingestion tool can move data between. For product teams that want open source, the three catalogs that matter in 2026 are Airbyte (700+ connectors by its own count, platform under the Elastic License 2.0), Meltano Hub (Singer taps and targets for 600+ sources and destinations, with MIT-licensed Meltano) and dlt (an Apache 2.0 Python library with verified sources and a generic REST API source). Fivetran, with 750+ sources, is the managed, closed-source benchmark they get compared against. Counts and licenses below come from each vendor's own pages, checked in October 2026.
Key Takeaways
- •Airbyte has the largest open catalog: 700+ connectors, with the platform under ELv2 and Airbyte's own connectors open source.
- •Meltano Hub indexes Singer taps and targets for 600+ sources and destinations; Matatika acquired the Meltano project after Arch shut down on March 1, 2026.
- •dlt is code-first: an Apache 2.0 Python library where verified sources and a REST API source stand in for a GUI catalog.
- •Fivetran lists 750+ sources and a Python Connector SDK, but it is not open source; it completed its merger with dbt Labs on June 1, 2026.
- •The catalog moves the data. Whether what lands is right is a separate job, and that is the job Data Workers owns.
| Catalog | Connector count (vendor's figure) | License | Custom sources | Deployment | Best fit |
|---|---|---|---|---|---|
| Airbyte | 700+ connectors | Platform ELv2; Airbyte's own connectors and CDK open source | Connector Builder and CDK | Self-hosted or Airbyte Cloud | Teams that want the broadest open catalog with a UI |
| Meltano Hub (Singer) | 600+ sources and destinations | Meltano MIT; Meltano SDK Apache 2.0; each tap carries its own license | Meltano SDK for Singer taps and targets | Self-hosted CLI, pipelines as code | Engineering teams that run ELT as code in Git |
| dlt | About 30 verified sources plus a generic REST API source | Apache 2.0 | Python code; dltHub adds a managed runtime | A library inside your own Python jobs | Product engineers who load data from code |
| Fivetran (managed benchmark) | 750+ sources; activation into 200+ destinations | Proprietary | Connector SDK (Python) | Fully managed SaaS | Teams that want no ingestion infrastructure to run |
Airbyte
Airbyte describes its catalog as '700+ connectors and counting for data pipelines and AI agents' and labels each connector with a support level such as Certified, so you can tell vendor-maintained connectors from the long tail. The core platform is licensed under the Elastic License 2.0, which allows self-hosting and modification but not offering Airbyte to third parties as a hosted service. Airbyte states that its own connectors and the Connector Development Kit remain open source.
For sources the catalog misses, the Connector Builder and the CDK let a team build its own. Airbyte runs self-hosted or as Airbyte Cloud. If you already run it, our you're on Airbyte guide shows how Data Workers works next to it, and Airbyte vs Fivetran covers the head-to-head.
Meltano Hub and Singer
Meltano is an MIT-licensed, code-first data integration engine built on the Singer standard, where taps extract and targets load. Meltano Hub is its index of connectors for 600+ sources and destinations, and the Meltano SDK (Apache 2.0) is the usual way to write a new tap or target. Because Hub indexes community connectors, quality and license vary tap by tap, so check maintenance activity before you depend on one.
Ownership changed this year: Arch, the company that had backed Meltano, shut down on March 1, 2026, and Matatika acquired the Meltano open source project, which remains open source and community driven. See you're on Meltano.
dlt
dlt (data load tool) is an Apache 2.0 Python library: you write a pipeline in code and dlt handles schema inference, normalization and loading. Its verified sources, around 30 today including Salesforce, Stripe and Zendesk, are maintained by the dlt team and community and tested against the real API, and a core REST API source covers HTTP APIs that have no dedicated connector. dltHub, the company behind dlt, adds a managed runtime and lists Oracle, SAP HANA and IBM DB2 among the sources it reaches beyond GUI catalogs.
For product teams that already write Python, dlt replaces a connector catalog with a library: fewer moving parts, more code to own. See you're on dlt.
Fivetran, the Managed Benchmark
Fivetran is not open source, but it sets the bar the others are measured against: 750+ sources, activation into 200+ destinations, and a Python Connector SDK for custom sources. Fivetran and dbt Labs completed their merger on June 1, 2026, and in May 2026 Fivetran became steward of the Great Expectations open source community. See you're on Fivetran.
How to Choose
Count the connectors you need, not the catalog size. List your sources, check each one in each catalog, and look at who maintains it: a certified or verified connector behaves differently from a community one nobody has touched in two years. Then pick the operating model: a UI with a managed option (Airbyte), pipelines as code in Git (Meltano), or a library inside your own jobs (dlt).
Connector catalogs are also different from metadata catalogs such as DataHub and OpenMetadata, which inventory the data you already have rather than moving it. If that is what you need, see you're on DataHub and you're on OpenMetadata.
Where Data Workers Fits
Every catalog here moves data. None of them tells you whether what landed is right: a sync can succeed while a column you bill on arrives a third empty. Data Workers, the agentic data platform, owns that job. It checks the tables your syncs land in Snowflake, BigQuery and Postgres for nulls, key uniqueness and row counts, flags anomalies against the baselines your team records, traces a bad table through your dbt project to the models that depend on it, and routes the proposed fix to the named owner for approval. Your team keeps running the syncs. Airbyte, Meltano, dlt and Fivetran connect over their API or MCP server today, and Data Workers ships 50+ connectors for the warehouses, orchestrators and tools around them. See the Data-Agents Swarm.
Frequently Asked Questions
What is a connector catalog? It is the set of prebuilt source and destination connectors an ingestion tool ships, such as Airbyte's 700+ connectors or the Singer taps and targets indexed on Meltano Hub.
Is Airbyte fully open source? The Airbyte platform is licensed under the Elastic License 2.0, which is source-available rather than OSI-approved open source. Airbyte states that its own connectors and its CDK remain open source.
Is Meltano still maintained? Yes. After Arch shut down on March 1, 2026, Matatika acquired the Meltano open source project, which remains open source and community driven.
When should a product team pick dlt over Airbyte? When the team already writes Python and wants ingestion inside its own code and jobs rather than a separate platform with a UI.
Is Fivetran open source? No. Fivetran is a managed, proprietary service with 750+ sources; its Connector SDK lets you build and deploy custom Python connectors on Fivetran.
Does Data Workers replace my connector catalog? No. Your ingestion tool keeps moving the data; Data Workers checks what lands, traces problems through dbt and gets the fix approved by the owner.