Prefect pricing guide details
Pricing Overview
Prefect operates on an open-core pricing model. The core workflow orchestration engine is fully open-source under the Apache 2.0 license, making it free to self-host with zero licensing fees. For teams that want managed infrastructure, Prefect offers two commercial platforms: Prefect Cloud for workflow orchestration and Prefect Horizon for managed AI infrastructure (MCP servers).
Prefect Cloud publishes its prices. The Hobby tier is free forever, Starter is $100/month, Team is $100 per user per month, and Enterprise is quoted. The figures are only half the story, because every paid tier bills against two limits that matter more than the headline: how many deployments you may register, and how many hours of Prefect Serverless compute are included.
The open-source edition supports unlimited flows, tasks, retries, and integrations with dbt, Kubernetes, and Docker. With over 21,800 GitHub stars and 10.4 million monthly downloads, it is a widely adopted Python orchestration framework. Moving to Prefect Cloud adds autoscaling workers, enterprise SSO, SOC 2 Type II compliance, and full observability dashboards. Prefect Horizon, their newer platform, provides an MCP gateway, server registry, and governance features for AI agent deployments; Prefect does not publish prices for Horizon, so that platform remains quote-driven.
Plan Comparison
Prefect Cloud (Workflow Orchestration)
| Feature | Hobby | Starter | Team | Enterprise |
|---|---|---|---|---|
| Price | Free forever | $100/month | $100/user/month | Custom |
| Users included | 2 | 3 | 4 to 8 | Custom |
| Deployments | Up to 5 | Up to 20 | Up to 100 | Custom |
| Prefect Serverless | 500 minutes | 75 hours | 225 hours | Custom |
| Run retention | 7 days | Extended | Extended | Custom |
| Compute model | Managed | Bring your own | Bring your own | Bring your own |
| Service accounts | - | - | Yes | Yes |
| Audit log | - | - | 24 hours | Extended |
| Workflow observability | Yes | Yes | Yes | Yes |
| Logging and alerting | Yes | Yes | Yes | Yes |
Two details in that table decide most bills. Starter is a flat monthly fee covering three users, while Team is charged per user, so adding a fourth person changes the billing basis rather than adding one seat to the same plan — the step from Starter to Team is the largest in Prefect's published pricing. And from Starter onwards you bring your own compute, so the plan fee buys the control plane and a Serverless allowance, not the machines your flows run on.
Prefect Horizon (AI Infrastructure)
| Feature | Free | Pro | Enterprise |
|---|---|---|---|
| MCP server deployments | Limited | Higher limits | Unlimited |
| MCP Gateway | Yes | Yes | Yes |
| Server Registry | Basic | Full | Full |
| Governance controls | - | Standard | Advanced |
| Custom domains | - | - | Yes |
| Priority support | Community | Standard | Dedicated |
Prefect does not publish dollar amounts for Horizon. Teams evaluating it should request a quote rather than assume it follows Prefect Cloud's published rates.
Hidden Costs and Considerations
Worker compute is always separate. Even on Prefect Cloud, flow execution happens on your own infrastructure from the Starter tier upwards. Prefect Cloud manages the control plane (scheduling, observability, triggers), and the Serverless allowance covers a fixed number of hours; beyond that, compute for running flows is billed through your cloud provider. Your total cost is the plan fee plus your cloud spend.
The Serverless allowance is the real ceiling. Hobby includes 500 minutes, Starter 75 hours and Team 225 hours. A team running hourly flows that each take four minutes consumes roughly 50 hours a month, which fits Starter; the same flows at five-minute granularity do not. Model your actual run frequency against the allowance before choosing a tier.
Deployment count is a second ceiling. Five deployments on Hobby and twenty on Starter are reached quickly by teams that register one deployment per environment, because a pipeline promoted through dev, staging and production counts three times.
Self-hosting infrastructure costs. The open-source Prefect server runs anywhere, but production deployments need a PostgreSQL database, a process to run the Prefect API server, and compute for workers. These costs are paid directly to your cloud provider and scale with concurrent flow runs and data volume.
Migration and onboarding effort. Teams moving from Airflow need to rewrite DAGs as Python functions with Prefect decorators. Migrating hundreds of existing DAGs represents meaningful engineering time; plan for a transition period where both systems run in parallel.
Connector and integration costs. Prefect integrates with dbt, Kubernetes, Docker, Snowflake and dozens of other tools. The integrations are free and open-source, but the underlying services carry their own costs.
Multi-platform licensing. Organizations using both Prefect Cloud and Prefect Horizon may face separate billing for each platform. Clarify with Prefect sales whether bundled pricing is available.
How Prefect Pricing Compares
Prefect sits in a different pricing category than most data pipeline competitors. Tools like Stitch (starting at $25/month), Hevo Data (starting at $25/month) and Airbyte Cloud (starting at $10/month) charge primarily for data movement volume — rows or connectors. Prefect charges for orchestration control plane features and a compute allowance. Direct comparison is difficult because Prefect often runs alongside these tools rather than replacing them.
| Tool | Model | Starting Price | Free Tier | Primary Billing Unit |
|---|---|---|---|---|
| Prefect | Open Core | $100/month | Yes (self-hosted + Hobby) | Deployments / Serverless hours |
| Stitch | Freemium | $25/month | Yes (1 user) | Rows replicated |
| Hevo Data | Freemium | $25/month | Yes (1M rows) | Events processed |
| Airbyte | Freemium | $10/month | Yes (self-hosted) | Connectors / credits |
Prefect's entry price is the highest of the four, and that comparison is less lopsided than it looks: $100/month buys orchestration for an entire platform, while $25/month at Stitch buys a row allowance that a single busy connector can exhaust. The two are billed on different things.
Prefect's strongest cost advantage remains the fully open-source self-hosted option. Teams with existing Kubernetes clusters and DevOps capacity can run Prefect at zero software cost, paying only for their own compute. Neither Stitch nor Hevo Data offers self-hosting at all.
The 73% cost reduction reported by Endpoint, switching from Astronomer to Prefect Cloud, suggests meaningful savings for teams migrating from commercial Airflow distributions, though results depend on workflow volume and infrastructure configuration. Cash App also reported faster deployment velocity after adopting Prefect Cloud.
For organizations evaluating Prefect against managed Airflow providers, the trade-off is between Prefect's Python-native developer experience with published, seat-and-allowance pricing, and Airflow's larger ecosystem with per-environment managed costs. The right choice depends on team size, existing infrastructure and whether the self-hosted path is viable for your operations team.