Pydantic BaseModel tool parameters not supported in mcp-lambda-handler (all params become string)?

Question

Can I use pydantic `BaseModel` classes as tool parameters with the AWS MCP Lambda handler from `awslabs/mcp`? When I call `tools/list`, all my tool parameters are serialized as `string` with no descriptions, and I cannot get automatic validation or JSON schema generation for complex inputs.

AgentArmorHQ team · Accepted Answer

This is **not supported**: as of the closing date, `mcp-lambda-handler` in `awslabs/mcp` only accepts basic Python types (`str`, `int`, `float`, `bool`) as tool parameters. Pydantic `BaseModel` subclasses are not detected for schema generation or validation, so complex parameters are not passed through as validated model instances. The feature request was closed as **NOT_PLANNED** after being marked stale, and no maintainer fix or official workaround was provided in the thread.

AgentArmorHQ team · Answer

**What @benitezfede tried on GitHub** (Worked)

A commenter who hit the problem hardcoded the response of `tools/list` so the tool schemas were described correctly, because `MCPLambdaHandler` was returning every parameter as `string` and without descriptions.

Source: https://github.com/awslabs/mcp/issues/684#issuecomment-3255885568

Pydantic BaseModel tool parameters not supported in mcp-lambda-handler (all params become string)?

2 answers