Data orchestration tools schedule the steps of a data pipeline, retry failures and keep a record of what ran.
This guide ranks thirteen of them on setup effort, the bill once runs multiply, how workflows leave the product and whether you can operate it without a reseller. Billing meters differ widely: users, credits, actions, minutes, hours or instances.
Vendor money can buy visibility on this page and nothing else. No entry is softened, criticism is kept, and a click on a vendor link brings us no revenue. How that works.
In short
What data orchestration tools software does
Data orchestration software coordinates tasks across extraction, transformation and loading systems with scheduling, dependencies, retries, alerts and run history, run by a vendor or as your own engine.
Five tests decide the order: setup effort, real cost, how your data leaves, whether you can walk away, and who the data orchestration tool was built for. On top of them sits one rule. The top three places are for challengers, which means vendors outside the big names, where the category has any, and the household brands come in from fourth place. For why these five were chosen, and why nothing here gets a numeric score, see how we work.
YAML-defined workflows on an Apache-2.0 engine with no cap on flows
Ranked #1 of 13 in Best Data Orchestration Tools in 2026.
Open sourceSelf-hostablePricing on requestElsewhere
The pricing page splits Kestra into an open-source edition, Enterprise and Cloud. Both editions list unlimited flows and executions and more than 2,400 plugins, while the Enterprise column adds SSO, LDAP, SCIM, RBAC, a secrets manager, worker groups and approvals that pause an execution.
The company says the project began in 2020 at Leroy Merlin France to replace Airflow. Cloud is a managed service offered through a request-access form.
What stands out
Apache-2.0
Per instance
Self-hosted or managed
Where it costs you
The open edition runs its queue on PostgreSQL or MySQL, while Kafka, Redis and AMQP queues are listed under Enterprise.
Flows are written as YAML, which asks teams used to Python-only definitions to learn a second authoring style.
Right for
Platform groups that want orchestration as reviewable files, backed by a French engineering base
Wrong for
Buyers who want amounts visible before any sales call
France / United StatesOpen source engine; Enterprise sold per instance on annual subscription, quoted per organisation; Cloud by access request
Flat-priced pipeline orchestration from London that triggers dbt and warehouse jobs
Ranked #2 of 13 in Best Data Orchestration Tools in 2026.
Free tierPublished pricingEurope
Orchestra counts tasks, tokens and compute minutes per day instead of runs: 1,000 tasks, 100,000 tokens and 100 compute minutes on each self-service plan, with pipelines unlimited.
Its solution pages describe triggering Fivetran, Snowflake, Databricks and dbt jobs from one place. The terms name a London registered office and English law. Enterprise is a fixed cost per workspace rather than per user, and no amount is shown for it.
What stands out
Flat monthly fee
dbt and Snowflake focus
Serverless runs
Where it costs you
Included compute covers dbt and Python runs only, at 100 minutes a day on Lean and Scale-up, so heavy transformations lead to an Enterprise quote.
Scale-up stops at ten users and two environments, which a platform team with several staging stages outgrows quickly.
Right for
Lean analytics teams running dbt and warehouse jobs who want a predictable monthly bill
Wrong for
Organisations that need unlimited compute on a fixed plan
United KingdomFree Lean plan; Scale-up at a flat monthly fee; Enterprise custom; prices published
Scripts in 20+ languages chained into flows, self-hosted or hosted by Windmill
Ranked #3 of 13 in Best Data Orchestration Tools in 2026.
Open sourceSelf-hostablePublished pricingNorth America
Windmill treats a flow as versioned code with a scheduler, retries, suspend-and-resume and approval steps around it, self-hosted if you want. Nothing is trapped: the scripts run anywhere Python or TypeScript runs.
The price of that independence is that every integration begins with reading an API document and writing the calls yourself, so the time-to-first-flow is days rather than the afternoon Zapier or Make would take.
What stands out
AGPLv3 and Apache-2.0 mix
Per builder seat
Self-hosted
Where it costs you
No vendor-built connectors; every API call is your code
Business users cannot build in it unaided
Right for
Engineering teams that want orchestration without giving up their code
Wrong for
Business teams connecting SaaS tools without developers
United StatesOpen source Community edition at no cost; Enterprise per builder seat plus per worker; prices published
Python, SQL, R and dbt blocks in one editor, billed per minute
Ranked #4 of 13 in Best Data Orchestration Tools in 2026.
Published pricingNorth America
Mage's plan table lists Starter at one seat and one development environment, Team at up to ten seats with one environment included, and Enterprise with unlimited seats by contract.
Larger compute sizes use proportionally more execution minutes. AI credits are capped at 50 a month on Starter and 250 shared on Team. The code repository carries Apache-2.0, and the terms name a California corporation.
What stands out
Apache-2.0 code
Per execution minute
Monthly AI credits
Where it costs you
Billing meters execution minutes, so a pipeline waiting on a slow API costs more than one that finishes quickly.
Starter allows one developer seat and one development environment, so a second engineer means moving to Team.
Right for
Small data groups that prefer a visual editor with code blocks over config files
Wrong for
Buyers who want no per-minute metering on a managed plan
United StatesStarter and Team monthly fees plus a charge per execution minute; Enterprise on contract; prices published
Kubernetes-native workflow engine where every step runs as a container
Ranked #5 of 13 in Best Data Orchestration Tools in 2026.
Open sourcePricing on requestElsewhere
The project page describes a container-native engine for orchestrating parallel jobs on Kubernetes, with workflows modelled as a sequence of tasks or as a DAG. It names machine learning, data processing and CI/CD pipelines as uses.
Argo is cloud agnostic and runs on any Kubernetes cluster. No price exists because the project is open source, so the real cost is cluster capacity and the engineers who maintain it.
What stands out
Apache-2.0
Kubernetes CRD
DAG and step layouts
Where it costs you
The engine runs only on a Kubernetes cluster, so a team without one must build that platform before the first workflow.
The project page names neither a support plan nor a managed offer, which leaves upgrades and on-call duty with the buyer.
Right for
Platform engineers who already run Kubernetes and want container-per-step pipelines
Wrong for
Analysts without cluster access or a platform team
Not stated by the vendorOpen source project; no vendor price list, cost is the Kubernetes cluster you run it on
Workflow engine with a free Hobby tier and per-user Team pricing
Ranked #6 of 13 in Best Data Orchestration Tools in 2026.
Free tierPublished pricingNorth America
Prefect's plan table gives each tier a different ceiling: 5 deployments on Hobby, 20 on Starter, 100 on Team and unlimited on Enterprise, with 500 serverless minutes, 75 hours and 225 hours included on the first three.
Starter lets you bring your own compute. The GitHub repository carries Apache-2.0. The site also announces that Prefect is acquiring Dagster Labs and expects to operate as one company under the Prefect name from August 2026.
What stands out
Apache-2.0
Per user on Team
Acquiring Dagster
Where it costs you
Run retention is 7 days on Hobby and Starter and 14 days on Team, which is short for audits that look back over months.
Team is billed per user within a four to eight user range, so cost grows with headcount before pipeline volume does.
Right for
Small data groups that want a hosted scheduler and can add governance later
Wrong for
Organisations needing SAML or OIDC login but not an Enterprise contract
United StatesFree Hobby tier; Starter flat monthly; Team per user per month; Enterprise custom; prices published
Asset-based orchestration billed per credit, now joining Prefect as a product
Ranked #7 of 13 in Best Data Orchestration Tools in 2026.
Published pricingNorth America
On the pricing page one credit equals one asset materialisation or one op executed. Solo is $10 a month plus $0.040 per credit for one user, Starter is $100 plus $0.035 for up to three users, and Pro is quoted.
A 30-day trial opens every feature. The open-source repository is Apache-2.0, and the terms name Elementl, Inc. as provider under California law.
What stands out
Apache-2.0
Per credit
Asset-based
Where it costs you
Every op execution counts as a credit, so a job split into many small ops can cost more than the same logic as one asset.
SAML, SCIM provisioning, audit logs and EU data residency are listed against Pro, which is sold by contact only.
Right for
Teams that model pipelines as data assets and can estimate monthly credit volume
Wrong for
Teams that would rather wait until the Prefect merger completes
United StatesSolo and Starter monthly fee plus a per-credit rate; 30-day trial; Pro quoted; prices published
Managed Apache Airflow billed by deployment size and worker hours
Ranked #8 of 13 in Best Data Orchestration Tools in 2026.
Published pricingNorth America
Astronomer's page lists Developer deployments from $0.35 an hour, Team from $0.42, workers from $0.13 an hour in seven sizes, and dedicated single-tenant clusters from $2.40 an hour. Standard multi-tenant clusters carry no cluster charge on any plan.
Airflow itself is Apache-2.0 in the Apache repository. The terms name Astronomer, Inc. under New York law, and Business Critical adds cross-region disaster recovery with a 30-minute support SLA.
What stands out
Apache Airflow
Metered by the hour
Private Cloud option
Where it costs you
Deployments run continuously at a fixed hourly price, so an idle scheduler still bills while the workers scale to zero.
Audit logs keep 7 days on Team and 90 on Business, and SSO enforcement starts at Business, which is a quote.
Right for
Teams committed to Airflow DAGs who want someone else to run the scheduler
Wrong for
Buyers who want one predictable monthly invoice for a small team
United StatesDeveloper and Team metered per hour; Business, Enterprise and Private Cloud quoted; prices published
Durable code-first workflow execution priced per action and gigabyte-hour
Ranked #9 of 13 in Best Data Orchestration Tools in 2026.
Published pricingNorth America
Temporal Cloud lists $50 per million actions, falling to $25 per million with self-service volume discounts, plus $0.042 per GB-hour of active storage and $0.00105 for retained storage.
New accounts receive $150 in credits for 90 days. Developer support costs 10 percent of usage, and Business support starts at $500 a month. The server repository is MIT-licensed, and the terms apply Washington law.
What stands out
MIT server
Per action
Eight SDKs
Where it costs you
Every workflow start, activity, timer, signal and query counts as a billable action, so chatty designs cost more than coarse ones.
The page lists SDKs but no warehouse or SaaS connectors, so extraction and loading steps must be written by your developers.
Right for
Engineering teams with long-running, failure-prone processes who already write Go, Java or Python
Wrong for
Analysts who expect a drag-and-drop pipeline builder here
United StatesPay-as-you-go per action plus storage; trial credits; support add-ons; prices published
Enterprise workload automation deployed on-premises, in containers or as SaaS
Ranked #10 of 13 in Best Data Orchestration Tools in 2026.
Self-hostablePricing on requestNorth America
Stonebranch groups its offer under Universal Automation Center, a service orchestration and automation platform, with event-driven automation, analytics, self-service automation and an Integration Hub.
The data pipeline page promises centrally managed workflows across end-to-end pipelines, plus jobs-as-code. Deployment can be on-premises, containerised or SaaS-based. The contact page lists Alpharetta as headquarters and a Frankfurt GmbH for EMEA, each with a support number.
What stands out
Universal Automation Center
Jobs-as-code
On-premises or SaaS
Where it costs you
Only a demo button stands where a price would be, so no two vendors can be set side by side before a sales conversation.
Its pages describe IT automation across hybrid environments first, so a pure analytics team may pay for scope it never uses.
Right for
Operations teams that run batch jobs, file transfers and pipelines on one scheduler
Wrong for
Start-ups wanting instant self-service sign-up with amounts shown
BMC's workload scheduler for batch and data pipelines, as SaaS or installed
Ranked #11 of 13 in Best Data Orchestration Tools in 2026.
Pricing on requestNorth America
BMC's Control-M page lists workflow automation, AI workflow orchestration, data pipeline orchestration, supply chain and telecommunications workflows as use cases, and the header offers Free Trials and Demos next to Request Pricing.
A Jobs-as-Code path is listed under developers, and the integrations shown include SAP Integration Suite, Dataiku and Azure AI Foundry. The page says the product is available as SaaS or on premises.
What stands out
BMC Software
SaaS or on premises
Jobs-as-Code
Where it costs you
The product page carries a Request Pricing button and no amounts, so a budget figure starts with a call to BMC.
The product spans job scheduling, file transfer and business workflows, which is more than a pipeline-only team normally needs.
Right for
Large organisations already scheduling batch work that want data pipelines in the same console
Wrong for
Small teams wanting a code-first engine they can inspect
United StatesRequest pricing through sales; trials and demos offered
AWS state-machine service priced per state transition or per request
Ranked #12 of 13 in Best Data Orchestration Tools in 2026.
Free tierPublished pricingNorth America
The pricing page states $0.000025 per Standard state transition in US East (N. Virginia), and its worked example of 100,000 executions of a four-step machine comes to $9.90 a month after the 4,000-transition free tier.
Express Workflows count requests and a duration rounded up to the nearest 100 ms, plus the memory used. Nothing on the page prices a seat or a user.
What stands out
Per state transition
Express workflows
AWS service
Where it costs you
A state transition is billed each time a step runs, retries included, so error-prone pipelines raise their own bill.
The pricing page and its examples sit entirely within AWS, so orchestrating jobs on another cloud needs your own integration code.
Right for
Teams whose pipelines already run on AWS services and who accept per-step billing
Wrong for
Buyers who want one tool across several clouds
United StatesStandard priced per state transition, Express per request and duration; free tier; prices published
Azure pipelines billed per activity run and integration runtime hour
Ranked #13 of 13 in Best Data Orchestration Tools in 2026.
Self-hostablePricing on requestNorth America
Data Factory wins on procurement and on reach: it is already in the agreement, and the self-hosted integration runtime pulls from databases behind the firewall without opening anything to the internet. That combination is hard to beat for a hybrid estate.
Day to day it is unpleasant. Pipelines are configured through a slow interface, failures are investigated in run histories, and because each activity run is billed, teams cram logic into fewer activities and make the pipelines harder to follow.
What stands out
Pay-as-you-go
Integration runtime
Azure service
Where it costs you
Authoring and debugging experience is poor
Per-activity pricing distorts pipeline design
Right for
Azure estates moving data between on-premise systems and the cloud
Wrong for
Teams needing a broad catalogue of SaaS application connectors
United StatesPay-as-you-go per activity run and integration runtime hour; rates vary by region
Data orchestration software coordinates tasks across extraction, transformation and loading systems with scheduling, dependencies, retries, alerts and run history, run by a vendor or as your own engine. Differences that change the outcome seldom sit in the features. Here is the sequence we would follow.
01
Decide whether you need a published price
5 vendors on this page sell by quote, while 8 publish a pricing model you can read today: Orchestra, Windmill, Mage, Prefect, Dagster, Astronomer, Temporal, AWS Step Functions.
02
Ask which country the contract runs through
Headquarters range over 4 countries and 3 regions: North America 10, Elsewhere 2, Europe 1. Legal exposure follows the vendor, not only the data centre, so read the contracting entity before you sign.
03
Consider whether you want the source
Open source here means Kestra, Windmill, Argo Workflows. The price of self-hosting is the time your team spends maintaining it.
How orchestration is metered: users, credits, actions, minutes or hours
Every vendor in this market counts a different thing, and the unit decides which design is cheap. Prefect bills Team per user, Dagster bills a credit for each asset materialisation or op, Temporal bills actions plus storage, Mage bills execution minutes, Astronomer bills deployment and worker hours, AWS Step Functions bills state transitions, and Orchestra caps tasks, tokens and compute minutes per day inside a flat plan.
A pipeline with fifty small steps, a nightly batch and an event-driven flow produce very different invoices under each meter. Before reading any price page, list your pipelines with their step counts, run frequency and waiting time on outside systems, then translate that list into each meter and request a written estimate.
Count the steps, retries and run minutes one typical pipeline produces each month
Check whether idle time is billed, for example continuous deployments or waiting execution minutes
Ask which meter the vendor raises first when your volume doubles
Open core and the features sold separately: SSO, audit logs and access control
Most open-source engines in this list keep governance in a paid tier. Kestra lists SSO, LDAP, SCIM, role-based access and audit logs under Enterprise, Prefect keeps SSO and SCIM for Enterprise, Dagster lists SAML, SCIM and audit logs against Pro, Windmill adds SAML and audit logs in its Enterprise edition, and Orchestra names Okta SSO and multiple workspaces as Enterprise controls.
A team that starts on the free edition and then needs login through an identity provider usually meets a quote, not a price list. Decide on day one whether auditors or customers will ask who changed which workflow, because that single requirement often moves a shortlist from the entry plan to a negotiated contract.
List the identity and audit requirements before you read the entry plan
Check which plan holds SSO, SCIM, audit logs and per-project permissions
Ask what retention applies to audit logs and run history on the plan you would buy
Scheduler, asset model or durable runtime: what each product expects you to write
These products start from different ideas. Astronomer runs Apache Airflow, so work is organised as DAGs of tasks. Dagster models pipelines as data assets with checks and lineage. Temporal runs ordinary code in eight languages and keeps its state across crashes, but supplies no warehouse connectors. Argo Workflows runs every step as a container on Kubernetes.
Control-M and Stonebranch descend from batch job scheduling and add file transfer and pipeline features for operations teams. Kestra and Windmill sit between these camps, with YAML flows and multi-language scripts. Pick the model that matches how your engineers already think, because migrating the definitions later is the costly step, not the licence fee itself.
Write down which language and definition style your team will maintain
Check whether lineage, asset checks or only task logs come with the plan
Test one real pipeline end to end in the trial before comparing price
Ownership and history limits: the Prefect and Dagster merger and 7-day run logs
Independence has a concrete form here. Prefect announced on 13 July 2026 that it is acquiring Dagster Labs, and expects to operate as one company under the Prefect name from August 2026, while the Dagster site promises the name, open-source licence and current contracts remain.
Kestra, Orchestra and Mage are still independent companies on their own sites, and Temporal, Astronomer and Windmill list their own legal entities. Short history windows are the other trap: Prefect keeps run history for seven days on Hobby and Starter, and Orchestra counts daily tasks. Read the retention row and the termination clause, and ask how workflow definitions and run history leave if you move.
Ask whether an acquisition or merger is announced or pending
Check how many days of run history and logs the plan keeps
Confirm that flow definitions export as plain files or code without a licence
What goes wrong most often when buying data orchestration tools software
Picking a tool by its lowest tier, when Dagster credits, Temporal actions and Step Functions transitions all grow with pipeline complexity.
Assuming the open-source edition covers everything, then finding that SSO, audit logs or task runners belong to a paid tier.
Buying a job scheduler such as Control-M or Stonebranch for a three-person analytics group that only needs warehouse jobs run in order.
Ignoring ownership news, such as Prefect acquiring Dagster Labs, until a renewal arrives with changed terms.
07
Frequently asked questions
8 answers
What is the best data orchestration tools in 2026?
Kestra leads our ranking of 13. Kestra defines workflows in YAML and runs them on an Apache-2.0 engine, with no limit on flows or executions in either edition.
Engineering and product sit in Kestra Technologies SAS in France, and a US entity covers the Americas. Single sign-on, role-based access, audit logs and task runners belong to Enterprise, so shared use by several teams means buying that tier.
Which data orchestration tools publish their pricing?
8 out of 13 state a price model on their own site:
Orchestra: Free Lean plan; Scale-up at a flat monthly fee; Enterprise custom; prices published.
Windmill: Open source Community edition at no cost; Enterprise per builder seat plus per worker; prices published.
Mage: Starter and Team monthly fees plus a charge per execution minute; Enterprise on contract; prices published.
Prefect: Free Hobby tier; Starter flat monthly; Team per user per month; Enterprise custom; prices published.
Dagster: Solo and Starter monthly fee plus a per-credit rate; 30-day trial; Pro quoted; prices published.
Astronomer: Developer and Team metered per hour; Business, Enterprise and Private Cloud quoted; prices published.
Temporal: Pay-as-you-go per action plus storage; trial credits; support add-ons; prices published.
AWS Step Functions: Standard priced per state transition, Express per request and duration; free tier; prices published.
The other 5 quote per organisation.
Is there a free data orchestration tool?
Orchestra, Prefect, AWS Step Functions can be used for free, either as a free plan or a self-hosted edition.
Where are these data orchestration tool vendors established?
Across 4 countries (North America 10, Elsewhere 2, Europe 1):
Kestra: France / United States.
Orchestra: United Kingdom.
Windmill: United States.
Mage: United States.
Argo Workflows: Not stated by the vendor.
Prefect: United States.
Dagster: United States.
Astronomer: United States.
Temporal: United States.
Stonebranch: United States.
Control-M: United States.
AWS Step Functions: United States.
Azure Data Factory: United States.
Which data orchestration tools are open source?
Kestra, Windmill, Argo Workflows.
Which data orchestration tools can you host yourself?
Kestra, Windmill, Stonebranch, Azure Data Factory. The remaining 9 run only on the vendor's servers.
What should you use instead of Kestra?
The next in line are Orchestra, aimed at Lean analytics teams running dbt and warehouse jobs who want a predictable monthly bill, and Windmill, aimed at Engineering teams that want orchestration without giving up their code.
Who should not buy Kestra?
Buyers who want amounts visible before any sales call. The open edition runs its queue on PostgreSQL or MySQL, while Kafka, Redis and AMQP queues are listed under Enterprise..
Makers of data orchestration tools products can ask us to look at theirs. Say what it does and who uses it. We choose what to add, and nothing can be bought; see software advice for what a listing means.
