Server has 6 tools with complete input schemas and descriptions, but quality is inconsistent. All tools use action verbs (moveEyes, setEyelids, blink, lookAt, randomMovement, setAutoBlink), which is good. However, descriptions are terse (10-40 chars), below the 50-200 char baseline for LLM optimization. Parameter descriptions are present but minimal. No output schema documentation. Error handling returns isError flag but lacks recovery guidance. Tools are single-responsibility and well-composed. Schemas use Zod validation with proper type constraints (enums, min/max bounds), which is strong. Missing: pagination (not applicable here), per-tool output schema docs, and actionable error messages.

• Expand tool descriptions to 50-200 chars. Example: 'moveEyes: Move eyes to a specific horizontal (0-180) and vertical (0-180) position. Use this for precise positioning; use lookAt() for named directions (left, right, up, down, center).'
• Document output schema in tool descriptions or via a separate schema field. Example: 'Returns {content: [{type: "text", text: string}], isError: boolean}. On success, text describes the action taken. On failure, text contains the error message.'
• Add recovery guidance to error messages. Example: 'Failed to send command: Connection timeout. Verify ESP32 is running at 192.168.50.195:80 and reachable from this network. Retry the command.'
• Add tool annotations: mark all 6 tools with destructiveHint: true since they modify animatronic state. Consider idempotentHint: true for moveEyes and lookAt (repeated calls with same params produce same result).
• Enhance parameter descriptions with physical context. Example: 'horizontal: Horizontal servo position in degrees (0=far left, 90=center, 180=far right). Use lookAt() for named directions.'
• Add a discovery tool or resource that documents the ESP32 API endpoints (/move, /eyelids, /blink, /random, /autoblink) and their parameter formats, so agents understand the underlying capabilities.