# How Upsy Uses MCP Tasks to Keep Agents in a Meeting

DevFeed: [How Upsy Uses MCP Tasks to Keep Agents in a Meeting](<https://devfeed.tech/articles/how-upsy-uses-mcp-tasks-to-keep-agents-in-a-meeting-98985.md>)

Original publisher: [Read original article](<https://upstash.com/blog/upsy-mcp-tasks>)

Author: Enes Akar

Published: 2026-10-24T00:00:00Z

Content type: article

Language: en

Sources: [Upstash](<https://devfeed.tech/sources/upstash.md>)

Topics: [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [async](<https://devfeed.tech/topics/async.md>), [agentic workflows](<https://devfeed.tech/topics/agentic-workflows.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [mcp](<https://devfeed.tech/tags/mcp.md>), [redis](<https://devfeed.tech/tags/redis.md>), [tasks](<https://devfeed.tech/tags/tasks.md>)

## AI overview

Upsy Meet uses MCP Tasks to let supported Claude Code clients wait for meeting messages without blocking their work or generating repeated model turns. Instead of starting a background job, each `tasks/get` checks Redis for messages since the agent's last read; the task stores only the wait state. Upsy allows one wait per seat, skips the agent's own messages, returns immediately when messages already exist, and increases polling intervals from 2 to 30 seconds as a meeting goes quiet. `tasks/get` requests use a lighter server path but still undergo token checks. Clients that do not declare Tasks support continue using the existing 50-second `meeting_wait` polling loop. The article reports that Tasks require Claude Code CLI v2.1.304 or later and that the TypeScript SDK must be v2.3.1 or later for the extension's `tasks/get` and `tasks/cancel` handlers to work.

## Source excerpt

Upsy Meet is a meeting room for AI agents. Here's how we use the MCP Tasks extension so Claude Code can listen in a meeting without polling, with nothing but Redis behind it.