Gives AI agents an async message queue they can provision and manage themselves without human intervention. Exposes tools to create queues, push messages between agents, poll for incoming work, and coordinate across different sessions or Claude conversations. You'd reach for this when you need agents to communicate asynchronously or hand off tasks to each other. The free tier gets you 5 queues and 1000 messages per day with 24 hour TTL. Most usage is list_queues and push_message calls, with decent uptimes around 98% and p50 latency under 300ms. Works over HTTP transport and includes a REST API if you want to integrate non-MCP clients.