Standardize MCP Argument Schemas

[chris-sanders]

Summary

This PR implements standardized MCP argument schemas by removing the non-standard args wrapper that was automatically added by FastMCP. This change ensures compatibility with PydanticAI and other MCP clients that expect standard schema formats.

Changes Made

Core Implementation:

• Transformed all 6 tool functions to use direct parameters instead of Args classes
• Updated initialize_bundle: source, force, verbosity parameters
• Updated list_available_bundles: include_invalid, verbosity parameters
• Updated kubectl: command, timeout, json_output, verbosity parameters
• Updated list_files: path, recursive, verbosity parameters
• Updated read_file: path, start_line, end_line, verbosity parameters
• Updated grep_files: pattern, path, recursive, glob_pattern, case_sensitive, max_results, max_results_per_file, max_files, verbosity parameters

Test Updates:

• Updated all unit tests (187 tests pass)
• Updated all integration tests (51 tests pass)
• Updated all e2e direct tests (10 tests pass)
• Added schema validation regression test to prevent future args wrapper usage

Before and After

Before (Non-standard):

{
  "properties": {
    "args": {
      "$ref": "#/$defs/InitializeBundleArgs"
    }
  },
  "required": ["args"],
  "title": "initialize_bundleArguments",
  "type": "object"
}

After (Standard MCP):

{
  "properties": {
    "source": {"type": "string", "title": "Source"},
    "force": {"type": "boolean", "default": false, "title": "Force"},
    "verbosity": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "title": "Verbosity"}
  },
  "required": ["source"],
  "title": "initialize_bundleArguments", 
  "type": "object"
}

Tool Calls

Before (Non-standard):

{
  "name": "initialize_bundle",
  "arguments": {
    "args": {
      "source": "bundle.tar.gz",
      "force": false
    }
  }
}

After (Standard MCP):

{
  "name": "initialize_bundle", 
  "arguments": {
    "source": "bundle.tar.gz",
    "force": false
  }
}

Quality Assurance

• ✅ All code formatting (ruff format) passes
• ✅ All linting (ruff check) passes
• ✅ All type checking (mypy) passes
• ✅ 187 unit tests pass
• ✅ 51 integration tests pass
• ✅ 10 direct e2e tests pass
• ✅ Schema validation regression test added

Compatibility

This change ensures compatibility with:

• PydanticAI MCP clients
• Claude Code MCP integration
• Standard MCP protocol implementations
• Any MCP client expecting standard schema formats

All existing functionality is preserved while enabling broader MCP ecosystem compatibility.