Server has 15 well-named tools with complete input schemas and descriptions. Naming follows verb_noun convention (list_, create_, export_, etc.). All tools have descriptions (avg ~80 chars, within baseline 34-392). Input schemas are properly typed with parameter descriptions. However, output schemas are not documented, LLMs cannot predict return structure for chaining. Error handling is minimal; no recovery guidance or actionable error messages visible. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Parameter descriptions lack format constraints (e.g., file_path accepts any string; no validation guidance). Some complex params like 'feature' object in add_feature/update_feature lack schema detail.

• Add output schema documentation to every tool. Use JSON Schema to declare return type, fields, and field types. E.g., list_documents returns {type: 'array', items: {type: 'object', properties: {documentId: {type: 'string'}, name: {type: 'string'}, defaultWorkspaceId: {type: 'string'}}}}.
• Declare tool annotations in MCP schema: readOnlyHint=true for export_*, list_*, get_*; destructiveHint=true for delete_feature; idempotentHint=true for read-only tools. This enables agents to reason about side effects.
• Expand add_feature and update_feature 'feature' parameter schema. Define properties object with required fields (e.g., featureType, parameters). Link to OnShape API docs or provide a minimal example schema.
• Add error handling guidance to tool descriptions. E.g., 'import_cad may timeout if file is large; if timeout_s is exceeded, the import may still complete asynchronously, check document status separately.' Provide recovery paths.
• Document file path constraints in export_stl, export_step, import_cad descriptions. E.g., 'file_path must be an absolute path; relative paths are not supported. Directory must be writable by the server process.'