Server has 5 tools with explicit schemas and descriptions, but quality is uneven. All tools are READ_ONLY and lack destructive/idempotent annotations. Descriptions are present but brief (avg 95 chars, below 194-char baseline). Parameter descriptions are minimal or absent. No output schemas documented. Error handling is basic (generic error messages, no recovery guidance). Tools are well-named with action verbs (list_, get_, ask_, chat_result, bot_chats), but parameter documentation is sparse. The ask_bot tool has internal polling logic (60s timeout) but no guidance on timeout recovery. No pagination support despite potential for large result sets (bot_chats returns 'latest 100' but no limit/offset params). Missing field-level descriptions for complex params like 'history' array items.

• Add outputSchema to each tool definition documenting response structure. E.g., ask_bot should declare: {type: 'object', properties: {job_id: {type: 'string'}, status: {type: 'string', enum: ['done', 'pending', 'timeout']}, reply: {type: 'string'}}}.
• Expand parameter descriptions with format/constraint details. E.g., 'question' param: 'Visitor question (max 1000 characters, plain text)'. 'history' items: 'Array of {role: "user"|"assistant", content: string}'. 'botId': 'Numeric bot ID from list_bots response (integer, required)'.
• Add error recovery guidance to tool descriptions. E.g., ask_bot: 'If timeout occurs, call chat_result(job_id) to poll later. If job_id not found, the job has expired (TTL ~10 min).' chat_result: 'Returns {status: "pending"} if job still processing. Retry after 1-2 seconds.'
• Declare tool annotations in schema. Add 'readOnlyHint: true' to all 5 tools' inputSchema to signal safety. This enables agents to reason about side effects.
• Add pagination support to bot_chats. Include 'limit' (1-100, default 20) and 'offset' (default 0) params. Return {chats: [...], total: N, offset: M} so agents can iterate over large chat histories.