I apologize, but I need to be honest about what I can definitively say about this topic. The ECS worker's interaction with AWS's DescribeTasks operation is handled by the AWS ECS integration, but I cannot find authoritative documentation about the specific polling frequency. Here's what I can tell you: 1. The AWS ECS service itself has API rate limits for the DescribeTasks operation: - The AWS API throttling limit for DescribeTasks is 100 transactions per second (TPS) by default - This is a regional limit that can be adjusted by contacting AWS support if needed 2. For the Prefect ECS worker specifically: - The worker checks task status as part of its normal operation - However, I cannot make a definitive statement about the exact polling frequency without being able to verify it in the code If you need to know the exact polling frequency for your use case, I would recommend: 1. Using AWS CloudWatch metrics to monitor your actual API usage 2. Setting up AWS CloudTrail to track DescribeTasks API calls 3. If you're concerned about hitting limits, you can implement exponential backoff in your flow runs or contact AWS support to request a limit increase Would you like me to: 1. Research any specific aspect of ECS worker configuration that might be more relevant to your use case? 2. Look into general worker polling behavior in Prefect? Please let me know what would be most helpful for your situation.

Short answer: no — Prefect doesn’t ship an MCP server. What we do have: - Marvin (Prefect’s AI framework) has experimental support for connecting to MCP servers, so agents can use MCP tools. Docs: Marvin + MCP guide - If you want “MCP for Prefect” (e.g., control flows/runs via MCP), you’d need to build/host an MCP server that wraps the Prefect API or Python client. Prefect’s docs: Prefect Docs. Happy to help sketch that out. Are you looking to control Prefect via MCP, or to let an agent access external MCP tools while working with Prefect?