[Feature Request] pre- research for using tool via code execute

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

With the development of MCP and the need for agents to handle long-running tasks, agents typically preload many tools, which incurs significant token overhead. Is there a way to store tool descriptions and other information in the file system, allowing the agent to dynamically load tools and use them through code execution?

reference: https://www.anthropic.com/engineering/code-execution-with-mcp

Solution

No response

Alternatives

No response

Additional context

No response