Closes part of #539

Adds a client example demonstrating a pre-execution authorization callback pattern. Shows how to evaluate tool calls against a policy (allow / approval_required / deny) before execution — a common need when connecting agents to MCP servers at scale.

PR #1436 covers sampling examples. This is a complementary example focused on the tool-execution authorization loop.

Motivation and Context

The original issue requested client examples. Currently, there are no examples demonstrating how to evaluate tool calls against a policy before execution. This pattern is critical for secure agent architecture where some tools are safe to run automatically, but others require human-in-the-loop approval.

How Has This Been Tested?

Ran the example locally using uv run examples/snippets/clients/client_with_authorization.py. Verified that the client successfully connects to the mcpserver_quickstart subprocess, correctly allows the execution of safe tools (like add), and properly blocks/flags destructive tools.

Breaking Changes

None. This is purely an additive documentation/example change.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update

Checklist

• I have read the MCP Documentation
• My code follows the repository's style guidelines
• New and existing tests pass locally
• I have added appropriate error handling (N/A - Example script)
• I have added or updated documentation as needed

Additional context

We ran into the need for this exact pattern while building AgentIAM to manage tool authorizations at scale, and figured a clean, standalone example using dataclasses would be helpful for others trying to implement similar human-in-the-loop policies!