New: Cookbooks and AI ExplanationsStep-by-Step recipes to solve problems connected to Roadmaps and Cheat Sheets. Need more details? Use AI buttons for structured and simple explanations with concrete examples throughout the whole platform.Take a look
Your tool takes two minutes to finish, and the client gives up long before it's done.
What you'll have at the end
A tool that returns a task handle right away, and a second call that polls it to completed or failed
You need
A working MCP server already reachable by a client, with at least one tool whose real work can run past the time that client is willing to wait for an answer.
Not covered
Choosing a particular queue or workflow engine to run the background work itself, and pushing status updates to the client instead of having it poll for them.
Leans on
The isError flag that keeps a failed MCP tool call recoverable
for the single-call version of telling a handled failure from a crashed one, with no task in the picture at all.
Can your MCP server survive a load balancer?
once the task record lives in shared storage, the same routing and instance-recycling fixes keep a poll for that task landing on an instance that actually holds it.
Checked 18 Aug 2026
Part of the MCP Servers Implementation cookbook