Hi 👋
The global tool timeout that was added in 02c9694 is unfortunately not flexible enough, if you ask me. Even though it's a good first step.
We have a diverse range of MCP servers and their tools have different requirements for timeouts. Many tool calls should answer immediately, and very few should be able to wait very long. Think of a database MCP for Big Data. Queries sometimes need their fair share of time. Configuring a global timeout of 5 minutes might be okay, but it's raising the timeout for everything. Even things that don't need it and shouldn't have a possible timeout of 5 minutes.
We'd therefore appreciate the possibility to configure tool timeouts more granularly. The global timeout could still stick around, but maybe we can introduce something to more conveniently control individual tool timeouts that overrides the global timeout when configured.
Best,
Christoph
Hi 👋
The global tool timeout that was added in 02c9694 is unfortunately not flexible enough, if you ask me. Even though it's a good first step.
We have a diverse range of MCP servers and their tools have different requirements for timeouts. Many tool calls should answer immediately, and very few should be able to wait very long. Think of a database MCP for Big Data. Queries sometimes need their fair share of time. Configuring a global timeout of 5 minutes might be okay, but it's raising the timeout for everything. Even things that don't need it and shouldn't have a possible timeout of 5 minutes.
We'd therefore appreciate the possibility to configure tool timeouts more granularly. The global timeout could still stick around, but maybe we can introduce something to more conveniently control individual tool timeouts that overrides the global timeout when configured.
Best,
Christoph