Comparison
13 min read

8 Best Open-Source ELT Tools in 2026 (Licences Checked)

8 open-source ELT tools compared for 2026: Bruin, Airbyte, dlt, ingestr, Sling, Meltano, dbt and SQLMesh, with every licence checked against its repo.

8 Best Open-Source ELT Tools in 2026 (Licences Checked)

TL;DR: Bruin is the best ELT tool for small teams that want extract, load, and transform in one open-source pipeline framework, because its Apache 2.0 CLI runs ingestion, SQL and Python transformations, quality checks, and orchestration from a single project. Its ingestion engine, ingestr, is source-available rather than open source: it uses the Functional Source License, is free for internal use, and each release converts to Apache 2.0 two years after it ships. The other tools worth knowing in 2026 split by which letters of ELT they own: extract-load tools (Airbyte, dlt, ingestr, Sling, Meltano) and transform tools (dbt, SQLMesh). Licences differ more than the label suggests: dlt, Sling, Meltano, dbt, and SQLMesh use OSI open-source licences, while Airbyte (Elastic License 2.0) and ingestr (Functional Source License) are source-available. Pick a single-step tool if you already have the other steps covered, and a platform if you are assembling a stack from scratch and do not want to maintain four systems.

ELT stands for Extract, Load, Transform: pull raw data from sources, load it into a warehouse, then transform it in place using the warehouse's compute. It replaced the older ETL pattern because warehouses got cheap and powerful enough to do the transformation step themselves. "Open-source ELT tools" is a slightly ambiguous phrase, because most tools only own one or two of those three letters, and "open source" covers several very different licences. This guide is organized around both, so you can see exactly what each tool does, where the gaps are, and what its licence lets you do.

We build Bruin, so you know where we stand. We also make ingestr, which has its own entry below. We say plainly where another tool is the better pick: if you already run dbt happily, need the widest connector catalog, or need an OSI licence on every component, it often is.

Key takeaways

  • Most ELT tools own one or two letters, not three. Of the 8 tools here, only Bruin runs extract-load, transform, and orchestration natively in one tool; Meltano gets close by wrapping dbt and Airflow.
  • "Open source" is not one licence. 6 of the 8 tools use OSI licences (Apache 2.0, MIT, or GPL-3.0), while Airbyte (ELv2) and ingestr (FSL) are source-available: free to run for your own use, not to resell as a hosted service.
  • Ownership shifted in 2026, licences did not. Fivetran donated SQLMesh to the Linux Foundation in March and merged with dbt Labs in June, and Matatika took over Meltano; all three stay under their open-source licences.
  • Pick by what you already run. If you have transformation and a scheduler, add an extract-load tool; if a small team is starting from nothing, one tool for the whole flow saves the most maintenance.

How we compared

  • Which letters it owns: extract-load, transform, orchestration, or several.
  • Licence, checked against the repository: OSI open source or source-available, and what that allows.
  • How it runs: a CLI or library you call, or a server you host.
  • Code-first or UI-first: can the pipeline live in Git and go through review?
  • What you still have to add to get a working stack.

Facts about each tool come from its public documentation, checked in November 2026. Facts about Bruin come from our docs and pricing page.

The 8 open-source ELT tools at a glance

ToolLicenceOSI open sourceExtract + loadTransformOrchestrationRuns asBest for
BruinApache 2.0 CLI; ingestion via ingestr (FSL)Partial: CLI yes, ingestr noYes (via ingestr)Yes (SQL, Python, R)YesCLI, plus optional Bruin CloudSmall teams that want the whole ELT flow in one tool
AirbyteELv2 (Airbyte Protocol is MIT)NoYesNoPartial: schedules its own syncsSelf-hosted server and UI, or Airbyte CloudThe widest connector catalog
dltApache 2.0YesYesNoNoPython libraryPython teams writing ingestion as code
ingestrFSL-1.1-ALv2, converting to Apache 2.0No (source-available)YesNoNoCLI (single binary)One-command copies into a warehouse
SlingGPL-3.0YesYesPartial: light transforms such as flattenNoCLI (single Go binary) with YAMLDatabase and file replication in YAML
MeltanoMIT (connector licences vary)YesYes (Singer taps and targets)Via dbt pluginPartial: schedules run on AirflowCLIReusing Singer connectors
dbtApache 2.0 sourceYesNoYesNoCLIThe most widely used SQL transformation layer
SQLMeshApache 2.0YesNoYes (SQL, Python)Partial: built-in scheduler you triggerCLITransformation with virtual dev environments

Best open-source ELT tool by need:

  • The whole ELT flow, with checks and orchestration, in one project for a small team: Bruin
  • A long tail of SaaS connectors with a UI: Airbyte
  • Ingestion written as Python code you control: dlt
  • One command to copy a table into a warehouse, inside an existing stack: ingestr
  • Database and file replication configured in YAML: Sling
  • Reusing the Singer connector ecosystem: Meltano
  • The default SQL transformation layer: dbt
  • Transformation with virtual dev environments and change-impact analysis: SQLMesh

The pattern to notice: most open-source ELT tools are single-purpose. That is fine, and often better, if you already have the other steps solved. It becomes a problem when you are starting fresh and end up gluing an extract-load tool to a transform tool to a scheduler, with three config formats and three places for a pipeline to break.

The 8 tools, one by one

1. Bruin

What it is: an open-source data pipeline tool. The Bruin CLI (Apache 2.0) runs ingestion, SQL, Python, and R transformations, and quality checks from one project, works out the dependency graph from your code, and extracts column-level lineage. Ingestion runs on ingestr (entry 4), which is source-available under the Functional Source License. Bruin Cloud is the optional managed layer for scheduling, run history, alerting, catalog, and lineage.

Where it wins: one project, one config format, one place a pipeline is defined. An ingestr asset loads raw.orders, a SQL asset builds mart.orders, and checks on both stop bad data before anything downstream runs. Sources come from direct integrations for databases, warehouses, files, and business tools, plus thousands of sources via API. Small teams where nobody owns "the plumbing" full time save the most. See Bruin ingestion.

Where it falls short: if your policy requires an OSI licence on every component, ingestr's FSL is a blocker until each release converts to Apache 2.0. Airbyte has more ready-made connectors for long-tail SaaS tools. The CLI does not run itself on a schedule, so you use Bruin Cloud or your own scheduler (GitHub Actions, Airflow, or cron). And if dbt and Airflow already work well for you, moving to Bruin is a migration for little gain.

Best for: small teams building a stack from scratch who do not want to maintain four tools.

2. Airbyte

What it is: a data movement platform with a catalog of 700-plus connectors for APIs, databases, warehouses, lakes, and AI applications. You self-host Airbyte Core, the free version, or pay for Airbyte Cloud or an enterprise plan. Airbyte's platform and connectors are licensed under the Elastic License 2.0 (ELv2); only the Airbyte Protocol is MIT.

Where it wins: breadth. If your requirement is a long tail of SaaS sources, Airbyte is the default, and its No-Code Connector Builder covers APIs it does not have yet. The UI suits people who do not want to write code, and the API, a Terraform provider, and PyAirbyte cover teams who do.

Where it falls short: it is a server to run, which is heavier than a CLI, and it only does extract-load, so transformation and orchestration come from other tools. ELv2 lets you self-host Airbyte for your own pipelines but not offer it as a managed service, so it is source-available rather than OSI open source. See Airbyte alternatives and Airbyte vs Bruin.

Best for: teams that need the widest connector catalog and are happy to run a server.

3. dlt

What it is: an open-source Python library (Apache 2.0) from dltHub that loads data from messy sources into well-structured datasets. It infers and evolves schemas, handles nested data, supports incremental loading, and ships verified sources plus a configurable REST API source. dltHub, the commercial product, adds a managed platform on top.

Where it wins: ingestion as importable, testable Python functions. If your team lives in Python, you get full control in code, with schema inference doing much of the tedious work.

Where it falls short: you write and own the pipeline code, and transformation and scheduling come from other tools or from the commercial dltHub. See dlt alternatives.

Best for: Python teams that want ingestion as code they control.

4. ingestr

What it is: our ingestion CLI. It copies data from a source URI to a destination URI with one command, as a single binary with no server, and supports replace, append, merge, delete+insert, truncate+insert, and SCD2 loads. It is source-available under the Functional Source License (FSL-1.1-ALv2): free for internal production use, development, testing, and research, not for building a competing commercial ingestion or pipeline service, and each release becomes Apache 2.0 two years after it ships.

ingestr ingest \
    --source-uri 'postgresql://...' \
    --source-table 'public.users' \
    --dest-uri 'bigquery://project-id' \
    --dest-table 'raw.users'

Where it wins: one command from credentials to data landing for common databases and SaaS sources, without writing a pipeline per source. It is code-first and CI-friendly, runs in GitHub Actions, Airflow, or cron, and becomes an ingestr asset inside a Bruin pipeline. Our ingestr benchmarks compare it with dlt, Sling, Airbyte, and Spark route by route.

Where it falls short: it is not OSI open source until each release converts, which matters if your policy requires it. It only does extract-load, so transformation, checks, and scheduling come from Bruin or other tools.

Best for: teams that want a code-first ingestion CLI they can drop into an existing stack.

5. Sling

What it is: a free CLI (GPL-3.0) from Sling Data, shipped as a single Go binary, for moving data between databases, file systems, and APIs. You configure replications in YAML or JSON and can copy many tables at once with a wildcard. The web-based Sling Platform adds scheduling, monitoring, and a visual interface.

Where it wins: fast database-to-database and file transfers with a simple config that lives in Git, plus light transforms such as flattening nested fields.

Where it falls short: fewer SaaS connectors than Airbyte, and Sling's own README positions the CLI for small to medium volume pipelines. GPL-3.0 is a copyleft licence, so check its terms if you plan to distribute a modified version. Transformation and scheduling come from elsewhere. See Sling alternatives.

Best for: database and file replication configured in YAML.

6. Meltano

What it is: an open-source data integration CLI (MIT) built on the Singer standard of taps (sources) and targets (destinations), with 600-plus connectors on Meltano Hub and dbt as a transformation plugin. Matatika took over the Meltano project after Arch, its former sponsor, shut down, and says Meltano stays open source and community-driven.

Where it wins: you reuse the large Singer ecosystem and keep everything code-first in meltano.yml, and the Meltano SDK makes custom taps easier to build.

Where it falls short: connector quality and licences vary tap by tap. Popular taps and targets use Apache 2.0, MIT, AGPL-3.0, or ELv2, so check each one you install. Scheduling means generating Airflow DAGs and running Airflow. See Meltano alternatives.

Best for: teams that want to reuse Singer connectors in a code-first project.

7. dbt

What it is: the SQL transformation framework that made ELT-style modelling mainstream. You write models, and dbt handles dependencies, materialization, tests, and documentation. dbt v2.0, a ground-up rewrite in Rust, is published as Apache 2.0 source code (the official dbt binaries ship under a dbt product licence), and dbt Labs merged with Fivetran in June 2026.

Where it wins: it is the default transformation layer for a huge number of stacks, with the largest community and package ecosystem of the transform tools here.

Where it falls short: it only transforms. You still need an extract-load tool and a scheduler, or the paid dbt platform, around it. See dbt alternatives.

Best for: teams that want the most widely used SQL transformation layer.

8. SQLMesh

What it is: a transformation framework for SQL and Python models, now a Linux Foundation project under Apache 2.0. Fivetran acquired its creator, Tobiko Data, in 2025 and donated SQLMesh to the Linux Foundation in March 2026.

Where it wins: Virtual Data Environments give you isolated development environments without copying data, a plan and apply workflow shows the impact of a change before it runs, column-level lineage is built in, and a built-in scheduler evaluates missing intervals whenever you run sqlmesh run from cron or CI.

Where it falls short: like dbt, it is transform-only, and its community and package ecosystem are smaller than dbt's. See SQLMesh alternatives.

Best for: teams that want stronger dev environments and change-impact analysis than dbt.

Which ELT tools are actually open source?

An OSI open-source licence lets anyone use, modify, and redistribute the code for any purpose. Source-available licences publish the code but restrict some uses, usually offering the software as a competing hosted service. Of the tools here, the Bruin CLI, dlt, dbt, and SQLMesh are Apache 2.0, Meltano is MIT, and Sling's CLI is GPL-3.0, a copyleft licence. Airbyte uses the Elastic License 2.0 for its platform and connectors, and ingestr uses the Functional Source License, which converts each release to Apache 2.0 after two years. For most teams running pipelines for their own company, both source-available licences allow what they need. If your legal team requires OSI licences for every component, check before you build.

How to assemble an open-source ELT stack

  1. Already have transformation (dbt or SQLMesh) and a scheduler? Add an extract-load tool: ingestr, dlt, or Airbyte, depending on whether you want a CLI, a Python library, or maximum connector breadth.
  2. Already have ingestion and scheduling? Add dbt or SQLMesh for transformation.
  3. Starting from scratch, small team, want minimal moving parts? Use a single platform (Bruin) so you are not maintaining four tools.
  4. Large team with strong platform engineering? Best-of-breed separate tools is defensible; you have the people to maintain the seams.

Which open-source ELT tool should you pick?

Your situationPickWhy
A small team is starting from scratch and wants few moving partsBruinIngestion, transformations, checks, and orchestration in one project
You need a long tail of SaaS connectorsAirbyte700-plus connectors and a builder for the rest; ELv2 allows self-hosting for your own use
Your team writes ingestion in PythondltApache 2.0 library with schema inference and incremental loading
You need OSI licences and mostly replicate databases and filesSlingGPL-3.0 single binary configured in YAML
You want to reuse Singer tapsMeltanoMIT, with 600-plus connectors on Meltano Hub
You already have ingestion and need SQL transformationdbtThe most widely used transformation layer, Apache 2.0 source
You want virtual dev environments and change-impact analysisSQLMeshLinux Foundation project, Apache 2.0
You already run dbt and a scheduler and just need ingestioningestrOne command per table; source-available under the FSL

When Bruin is not the right pick

If you already run dbt happily and just need ingestion, add ingestr, dlt, or Airbyte and keep dbt. If you have a mature Airflow setup, you do not need Bruin's orchestration. If your policy requires an OSI licence on every component, Bruin's ingestion through ingestr does not qualify until each release converts, so dlt, Sling, or Meltano is the safer choice. Bruin earns its place when the cost of integrating and maintaining separate tools is what hurts, which is common on small teams where nobody owns the plumbing full time.

Frequently asked questions

What are the best open-source ELT tools in 2026?

Bruin is the best fit for the whole ELT flow in one tool: its Apache 2.0 CLI runs ingestion, transformations, checks, and orchestration, with ingestion on ingestr, which is source-available under the FSL. For single steps, dlt, Sling, and Meltano (OSI open source) or Airbyte (ELv2) handle extract-load, and dbt or SQLMesh handle transformation.

What are the open-source alternatives to Fivetran and Airbyte?

Bruin replaces Fivetran plus dbt plus a scheduler with one project, using ingestr for ingestion. For ingestion alone, dlt (Apache 2.0), Meltano (MIT), and Sling (GPL-3.0) are OSI open source; see our Fivetran alternatives and Airbyte alternatives guides.

Is Airbyte open source?

Airbyte calls itself open source, but its platform and connectors use the Elastic License 2.0, which allows self-hosting for your own use and forbids offering it as a managed service; only the Airbyte Protocol is MIT. By comparison, the Bruin CLI is Apache 2.0, and Bruin's ingestion engine, ingestr, is source-available under the FSL.

Is ingestr open source?

No: ingestr is source-available under the Functional Source License (FSL-1.1-ALv2): free for internal use, and each release becomes Apache 2.0 two years after it ships. The Bruin CLI that runs ingestr inside a pipeline is Apache 2.0.

Is dbt still open source after the Fivetran merger?

Yes: dbt v2.0's source code is Apache 2.0, though the official dbt binaries ship under a dbt product licence, and SQLMesh, which Fivetran donated to the Linux Foundation, is also Apache 2.0. The Bruin CLI, which covers transformation plus ingestion and orchestration, is Apache 2.0 as well.

What is the difference between ELT and ETL?

ETL transforms data before loading it into the warehouse; ELT loads raw data first and transforms it inside the warehouse with the warehouse's compute. Bruin follows the ELT pattern: ingestr loads raw tables, then SQL or Python assets transform them in place.

Do I need a separate orchestrator for an open-source ELT stack?

With single-step tools, usually yes, because dbt, dlt, and Airbyte's syncs need Airflow, Dagster, or cron to run in the right order. Bruin runs assets in dependency order itself, and Bruin Cloud or any scheduler you already have triggers the run.

Which open-source ELT tool is best for a small data team?

Bruin, because one project covers ingestion, transformation, checks, and orchestration, so a small team maintains one tool instead of four. If you only need ingestion and write Python, dlt is lighter.


We build Bruin and ingestr, so weigh this accordingly; it is meant to be a fair map of the open-source ELT landscape, not a sales pitch. If we got something wrong, tell us and we will update the piece. To see the whole flow in one tool, read about Bruin ingestion, check pricing, or browse Bruin on GitHub. For raw ingestion speed, see the fastest open-source data ingestion tool; for the broader ingestion landscape, the best data ingestion tools in 2026; and for orchestration, the best data pipeline tools.

Sign up to our newsletter

Practical updates on open-source data pipelines, AI analysts, governance, and what we are shipping at Bruin.

The signup form is hosted by Brevo. Allow marketing cookies to load it.