The server defines 2 tools with complete input schemas using Zod and explicit JSON Schema generation. Both tools have descriptions (194 and 181 chars respectively), well within the ideal 10 - 1024 range. Parameters are typed and mostly documented. However, output schemas are entirely undocumented, the tools return raw JSON from OpenAI without guidance on what fields the LLM should expect. Parameter descriptions are present but vary in quality; some (like 'Conversation messages') lack actionable detail about format constraints. No output field documentation, no pagination guidance, and error messages are generic. The server lacks tool annotations (readOnlyHint, destructiveHint) and does not guide the LLM on recovery strategies when API calls fail.

• Add a documented output schema for each tool. At minimum, describe the structure: 'Returns an OpenAI Chat Completion object with fields: id, model, created, choices (array of {finish_reason, message{role, content}}), usage {prompt_tokens, completion_tokens}. For web_search requests, choices[].message.content contains synthesis and search_results field with web citations.'
• Add tool annotations in the CallToolRequestSchema handler. Example: when returning tool definitions, include `readOnlyHint: true` for both web search tools so the LLM knows these are safe to call repeatedly without side effects.
• Enhance error messages with actionable recovery guidance. Catch specific OpenAI API error codes (401 Unauthorized, 429 Rate Limited, 400 Invalid Model) and return error responses like: 'OpenAI API returned 401 Unauthorized. Verify OPENAI_API_KEY is set correctly and has web-search model access. Available models: gpt-4o-search-preview, gpt-4o-mini-search-preview.'
• Document parameter constraints in descriptions. Update 'messages' description to: 'Array of message objects. Each must have role (user|assistant|system|developer for gpt-4o-search-preview; user|assistant|system for gpt-4o responses) and content (string, max 2000 chars per message for search context). Example: [{role: "user", content: "What is the weather in San Francisco?"}]'
• Add a security note to both tool descriptions: 'Requires OPENAI_API_KEY environment variable set server-side. Do not pass credentials as parameters.'
• Document API limitations in descriptions. Example: 'Search context size (low|medium|high) affects result verbosity; high may exceed context windows on large responses. Default: medium. Results are truncated to latest OpenAI API response; for extensive searches, consider multiple calls with refined queries.'