Tools have basic descriptions and clear naming (verb_noun pattern), but lack critical schema documentation and parameter descriptions. Input schemas are visible in quickstart.py but minimal: only parameter names and types, no descriptions for individual parameters. Output schemas are completely undocumented. Descriptions are adequate (50-150 chars) but lack actionable context for LLM selection. No error handling guidance, no pagination support despite async operations, and no documentation of side effects or prerequisites.

• Document output schema for each tool. Example: submit_batch returns {batch_id: string, valid_url_count: int, invalid_urls: [string]}. Include all fields the LLM needs to chain calls.
• Add parameter descriptions with constraints. Example: 'urls: array of company domain names (e.g., example.com, not https://example.com). Invalid URLs are reported but do not block batch creation.'
• Add enum constraint to output_types parameter: ['parent_companies', 'pe_ownership', 'corporate_families', 'pe_portfolios']. Replace free-form string with validated choices.
• Enhance download_run description to include cost estimation and confirmation flow: 'First call charges credits (see check_credits first). Subsequent calls for the same run_id are free. Returns a 1-hour download link.'
• Add error recovery guidance to tool descriptions. Example: 'If polling times out after 20 minutes, call poll_until_complete again with the same run_id, the run continues in the background.'