Developers often complain about MCP servers requiring manual activation before each session, a process many consider inefficient. An engineer writing on DEV Community shared how the goose agent addresses this friction through its Extensions Manager. This feature lets goose dynamically discover and enable only the MCP servers needed for a specific task, rather than loading a full suite of tools upfront.

The author demonstrated this workflow using the Github MCP Server. Although the server was disabled in the extension manager, simply asking goose to perform a task requiring Git access triggered an automatic search. Goose identified the relevant server and requested the necessary authentication confirmation, allowing the user to proceed without pre-configuration.

This approach solves a common pain point in AI agent workflows: resource bloat and security overhead. By enabling servers on-demand, goose streamlines productivity for users managing multiple integrations. The feature works across both desktop and terminal interfaces, offering flexibility for different development environments.

As AI agents gain access to more external tools, dynamic discovery like this may become standard practice for maintaining performance and security.