mcp tools timeout after 30s with no way to configure it, is this hardcoded?
hit this yesterday when testing mcp sdk 0.5.0 with a database tool that runs slow analytics queries. the tool times out after exactly 30s and there's no config option to change it. looked through the sdk docs and source and the timeout seems hardcoded. this is broken for any tool that needs to run longer operations - database queries, file processing, external api calls that are legitimately slow. is there a way to configure this or do i need to fork the sdk? seems like a basic prod requirement
tested @modelcontextprotocol/server-filesystem 1.0.1 and @modelcontextprotocol/server-sqlite 0.3.0 yesterday and both have 30s hardcoded timeout in the sdk. no config option, just fails after 30s on long-running queries
can you share the exact sdk version and config? need to repro this on our setup
we hardcoded 120s timeout in our wrapper after hitting this. shipping next week.
Tested @modelcontextprotocol/server-sqlite 0.3.0 yesterday with queries that take 90+ seconds on large tables and hit the exact same 30s timeout with no config option. The error message just says "tool execution timeout" with zero context about how to extend it. We ended up wrapping the mcp server with a custom timeout handler that retries with chunked queries, but this feels like something the sdk should support natively. Is there a config option we're missing or is 30s actually hardcoded in the protocol spec?
can you share the exact sdk version? tested @modelcontextprotocol/server-sqlite 0.3.0 yesterday and confirmed 30s timeout is hardcoded in the sdk source at line 247. no config option exposed