No Glue. No Lambda. No EventBridge. No Step Functions. QueryFlow runs your entire ETL workflow locally on your Mac. $299.99/year flat, no per-job pricing, no DPU hours.
A typical cloud ETL setup at a mid-sized company looks like this: AWS Glue for the transformations ($500-$5,000/month in DPU-hours), Lambda for the orchestration glue ($50-$200/month), EventBridge for scheduling ($30-$100/month), S3 for staging ($50-$300/month), CloudWatch Logs ($100-$500/month), Secrets Manager for credentials ($20-$80/month). Total: $750 to $6,180 per month in AWS bills. Plus the engineering time spent maintaining the whole apparatus.
When you strip away the cloud abstractions, most AWS ETL setups are doing simple things: pull data from a warehouse with SQL, transform it with Python or SQL, write the result somewhere. The cloud infrastructure exists primarily to schedule and orchestrate — it doesn't add computational power that wasn't already in your warehouse. The compute happens in Snowflake or Redshift or wherever your source data lives. AWS is mostly the conductor, not the orchestra.
QueryFlow is a desktop app that schedules and orchestrates ETL pipelines on your Mac. The compute still happens in your warehouse — Snowflake runs the queries, the same as before. The difference is the scheduling layer, the field mapping logic, the error handling, and the run history all live in QueryFlow on your Mac instead of in AWS services. The result: identical functionality, identical compute cost (your warehouse charges either way), but no AWS orchestration bill.
This is the fair question — what happens if your Mac is offline at scheduled run time? QueryFlow's catch-up model handles missed runs automatically on the next wake. For analytical workloads where data freshness within a few hours is fine, this works perfectly. For 24/7 SLA-bound workloads where the Mac being offline for an hour is unacceptable, AWS-based scheduling is still the right architecture. Honestly evaluate which category your pipelines fall into.
Be straightforward about the trade-offs. You lose: shared team access to a single source of truth (QueryFlow runs per-Mac, so different team members run different schedulers), enterprise observability tooling like Datadog and PagerDuty integration, multi-region failover, fine-grained IAM permission models. For solo developers and small teams, these are non-issues. For enterprise data platforms with strict compliance requirements, AWS remains the right choice.
The realistic migration is not 'rip out AWS today.' It's 'pick one Glue job per week, recreate it in QueryFlow, run them in parallel for 30 days, then turn off the AWS version.' Most teams find 60-80% of their AWS ETL workflows are excellent candidates for QueryFlow. The remaining 20-40% (the truly enterprise-grade pipelines with strict requirements) stay in AWS. The total AWS bill drops to a fraction of what it was, even though some pipelines remain.
A small data team's typical AWS ETL bill: $1,500/month average = $18,000/year. Plus 5-10 hours/month of engineering time maintaining the AWS infrastructure = $15,000-$30,000/year in fully-loaded engineer cost. Total: $33,000-$48,000/year. Same team running 80% of pipelines through QueryFlow: $299.99/year for the QueryFlow license + ~$300/month residual AWS bill for the remaining 20% of pipelines = $3,900/year. The math is uncomfortable for AWS.
QueryFlow replaces the orchestration layer: AWS Glue (for visual ETL and Python transformations), Lambda (for scheduled job execution), EventBridge (for cron-style triggers), Step Functions (for pipeline sequencing), and parts of CloudWatch Logs (for run history and error logs). What you keep in AWS: S3 if you store data there (QueryFlow writes to S3), Secrets Manager if you use it (QueryFlow has its own Keychain-based credential store), and your data warehouse compute (Redshift, Aurora) which QueryFlow connects to.
QueryFlow's scheduler runs while your Mac is awake. Configure macOS Power Schedule (System Settings → Battery → Schedule) to wake your Mac at scheduled times if jobs must fire overnight. Alternatively, dedicate an always-on Mac mini ($600 hardware cost) as the ETL machine — many teams find this is cheaper than a year of AWS Glue. For pipelines requiring true 24/7 cloud durability, keep those in AWS.
Yes. Amazon S3 is a destination (writeback for query results and pipeline outputs). Amazon Redshift is both a source and a destination via the Redshift Data API with IAM authentication. You can run a Postgres-to-Redshift pipeline through QueryFlow, with the orchestration on your Mac and the data compute happening in AWS — best of both worlds.
QueryFlow's Python transformation layer is built on the local Python 3.12 runtime with pandas. For datasets that fit in memory (typically up to 10-50GB depending on your Mac's RAM), this is plenty. For multi-billion-row Spark transformations, you genuinely need cloud compute — AWS Glue, EMR, or Databricks remain the right tools for that scale. QueryFlow does not try to compete in that segment.
QueryFlow v1.5 does not currently support external API triggers — pipelines run on schedule or manual click only. For CI/CD-triggered pipelines (running on git pushes, deploy events, etc.), AWS Lambda or GitHub Actions remain the right tools. Programmatic pipeline triggering is on the QueryFlow roadmap.
14-day free trial. Recreate one Glue job in QueryFlow this week and measure the AWS bill that does not appear next month.