Prefect

Prefect develops a workflow orchestration platform that turns ordinary Python functions into production workflows using two decorators, @flow and @task. The system provides state tracking, automatic retries, caching, crash recovery, scheduling (cron, event-driven, and backfills), logging, and dependency/DAG visualization without requiring a domain-specific language, YAML, or pre-declared DAG structures. Because workflows are plain Python, they can branch, loop, and create tasks dynamically at runtime. The product is positioned for data pipelines, machine learning training and experimentation, durable background tasks, platform engineering across microservices, and orchestration of LLM/agent workloads with human-in-the-loop control.

The offering has two main forms: the open-source Prefect framework, licensed under Apache 2.0, which can be self-hosted with full VPC control; and Prefect Cloud, a managed, high-availability deployment running the same core engine and adding a managed control plane, enterprise authentication, SSO, autoscaling, and enterprise features. Prefect Serverless runs flows on managed compute, while a hybrid model places lightweight workers inside the customer environment so code and data remain in place while the control plane handles scheduling and visibility. Work pools decouple workflow code from execution infrastructure, supporting Docker, Kubernetes, ECS, Google Cloud Run, Azure ACI, Modal, and serverless targets. Enterprise attributes advertised include SOC 2 Type II, HIPAA and GDPR alignment, 99.99% uptime, and a 30-minute support response SLA.

Prefect also maintains FastMCP, an open-source framework for building Model Context Protocol servers that connect AI agents to tools and data, and Prefect Horizon, an MCP gateway providing a single entrypoint for MCP servers with authentication, authorization, and per-tool governance. A Prefect MCP server allows AI assistants such as Claude Code, Cursor, Codex CLI, and Gemini CLI read-only access to deployments, flow runs, task runs, and logs.

Founding story

Founded in 2018 by Jeremiah Lowin, whose experience as a PMC member of Apache Airflow and exposure to distributed tools such as Dask informed the idea that orchestration should be expressed in native Python decorators rather than rigid DAG definitions. Third-party coverage frames the founding insight as an unmet need for more robust, flexible orchestration able to handle the growing complexity and scale of modern data operations, and as a reaction to the operational overhead imposed by existing tools such as Apache Airflow.

Business model

Open-core: the Prefect orchestration engine is released as open source under Apache 2.0 and can be self-hosted, while Prefect Cloud is a commercial managed deployment of the same engine that adds a hosted control plane, enterprise authentication, autoscaling, and support. Tiered plans include a free entry point and enterprise plans with a 30-minute support response SLA.

Commercial revenue comes from Prefect Cloud subscriptions (managed orchestration with enterprise features such as SSO, autoscale, priority support), layered on a free open-source framework; the 2021 funding announcement described a product-led growth strategy converting open-source community usage into commercial adoption.

Traction

The open-source framework reports 14.3M+ downloads per month, 23.6k+ GitHub stars, and 425+ contributors, with a community described as nearly 30,000 engineers (the company site cites 25,000+). The company states over 100,000 mission-critical workflows are orchestrated every minute. The Prefect package is cited at 6M+ monthly and FastMCP at 103M+ monthly on the homepage. At the time of the 2021 Series B, Prefect Cloud usage was growing 130% per quarter since its February 2020 public launch and had surpassed 200 million tasks processed. Customer-reported outcomes include Snorkel AI running 1,000+ flows per hour with a 20x throughput improvement, WHOOP cutting production incidents by 75%, and reported cost reductions of 70% (Cash App) and 73% infrastructure savings cited in third-party coverage.

Latest developments

Prefect acquired Dagster Labs in July 2026, with the company stating that nothing changes for users of either product. Current product positioning extends orchestration beyond data and ML to AI agents, including Prefect Horizon (a secure MCP gateway with authentication, authorization, and per-tool governance), the FastMCP open-source MCP server framework, a read-only Prefect MCP server for AI coding assistants, and Prefect Serverless managed compute.