Workflow orchestration and AI infrastructure
Prefect
Prefect builds workflow orchestration software for data, machine learning and AI agents. Its open-source Python framework adds scheduling, retries, caching and recovery to workflows, while Prefect Cloud provides a managed control plane. Teams can use managed compute or execute code inside their own infrastructure. FastMCP helps developers build Model Context Protocol servers connecting agents to tools and data; Prefect Horizon adds hosting, a capability registry, tool-level access controls and agent interfaces.
Founded in 2018, Prefect is led by founder and CEO Jeremiah Lowin. Chris White, a day-one colleague and longtime CTO, became president in 2025. Its engineering approach separates durable results from individual workflow identities: configurable cache keys let independent workflows reuse computations, while locks coordinate access and rollback hooks support compensating actions after failures. In July 2026, Prefect announced plans to acquire Dagster Labs, committing to maintain Dagster and Dagster+. Dagster’s focus on defining data assets complements Prefect’s emphasis on executing workflows.
In January 2026, the company reported one million FastMCP downloads a day, with some version powering 70% of MCP servers. FastMCP was also incorporated into the official MCP SDK. In July 2026, Prefect said it had operated profitably over the preceding year.
2 talks
Newest firstThe AI Pivot: With Chris White of Prefect & Bryan Bischof of Hex
Chris White · Bryan Bischof · Brittany Walker
AI Engineer Summit 2023 · 35:16
2 speakers at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Start here
- The AI Pivot: With Chris White of Prefect & Bryan Bischof of Hex
Start with this discussion to understand how Marvin and Hex Magic fit into broader questions about adding AI to existing software.
Chris White · Bryan Bischof · Brittany WalkerAI Engineer Summit 2023
- Your MCP Server is Bad and You Should Feel Bad
Use this talk to examine how focused tool surfaces and agent-centered workflows shape MCP server design.
Jeremiah LowinAI Engineer Code 2025
Messages from the stage
Evaluating an AI product investment
The joint discussion considers hiring and measuring product-level returns alongside technical choices such as retrieval, vector databases, and orchestration for unreliable model APIs.
Designing beyond REST API exposure
Lowin frames MCP servers as products designed for agents. His talk pairs that design goal with pragmatic API bootstrapping and concrete runtime concerns, including Claude Desktop tool caching and asynchronous background tasks.
Affiliations reflect each recorded session, not necessarily current employment. The AI Pivot is a joint discussion with Chris White, Bryan Bischof, and moderator Brittany Walker.

