Solution for 1cb44dc5-1f6a-49ff-bc2e-fcf1b619ea7d: Update attribute access to snake_case (e.g. tool.input_schema, result.is_error,

## Cause v2 uses snake_case Python attributes with camelCase aliases only for construction and inbound parsing; default model_dump emits snake_case. ## Steps 1. Search codebase for camelCase MCP field access (inputSchema, isError, nextCursor, structuredContent). 2. Replace with snake_case attributes per migration table. 3. Audit model_dump() and model_dump_json() calls; add by_alias=True, mode='json' for wire output. 4. Run integration tests against a live MCP peer expecting camelCase JSON. 5. If migration is blocked, constrain dependency to mcp<2 temporarily. ## Expected result No AttributeError on Tool fields; serialized JSON uses camelCase keys (inputSchema) recognized by MCP clients and servers. ## Misleading approaches - Fixing only AttributeError sites while leaving model_dump() emitting snake_case silently. - Assuming wire JSON changed because only Python attribute names changed. ## Limitations - Constructor kwargs still accept camelCase aliases but attribute reads must use snake_case. - Separate validation rules in v2 may reject empty input_schema objects at wire serialization time. ## Unknowns - Full inventory of camelCase references in a large codebase without automated search. ## Evidence claims - (documented_platform_behavior) v2 maps inputSchema AttributeError to snake_case rename; use input_schema attribute. — https://raw.githubusercontent.com/modelcontextprotocol/python-sdk/f1b6589088534632fef92238ee9750951e3c0185/docs/migration.md - (documented_platform_behavior) model_dump() without by_alias emits snake_case keys; by_alias=True produces wire camelCase. — https://raw.githubusercontent.com/modelcontextprotocol/python-sdk/f1b6589088534632fef92238ee9750951e3c0185/docs/migration.md