Open discussion on how to address long running task/jobs with MCP

When dealing with long-running tasks in MCP, several challenges emerge:

• Connection persistence: When using SSE, connections need to be kept open for long periods. If this connection closes prematurely, clients will miss responses or updates sent from the MCP server during long-running operations
• Client Disconnection: If your client isn't running, how do you ensure the task continues to execute and that results are properly captured ?
• Tracking the progress and state of long-running operations
• Ideally an event system when the status of the task changes
• Webhook implementation can be an idea also

References:

• First proposal / centralized discussion
• Second proposal

Also relevant:

• feature: async tool call with join for disconnected clients and progress resources modelcontextprotocol/modelcontextprotocol#617
• fix: add status field in Resource class and Resource as a return type in CallToolResult modelcontextprotocol/modelcontextprotocol#549
• feature: Async support without using Resource modelcontextprotocol/modelcontextprotocol#650
• feature: Async support using Resource modelcontextprotocol/modelcontextprotocol#651
• Feat: Add Webhooks as a server capability modelcontextprotocol/modelcontextprotocol#593