Eight tools with consistent but minimal quality. All tools have basic names starting with action verbs (list_, get_) and brief descriptions (17-78 chars). Input schemas are present and properly typed for all tools, but parameter descriptions are sparse or missing entirely. Output schemas are not documented. No error handling guidance, no field chaining for composition, and descriptions lack context for LLM tool selection. This is a functional but subpar definition set typical of internal beamline tools that prioritize API simplicity over agent usability.

• Expand all tool descriptions to 100 - 200 characters, following pattern:tool-description. State WHAT (retrieves motor positions), WHEN (call before issuing motion commands), ALTERNATIVES (get_motor_status for full info including goal), PRECONDITIONS (motor list must be retrieved via list_motors first).
• Add 'format' and 'constraint' to every parameter description. For 'channels' param: 'List of valid AI channel names (strings). Valid names: [description of discovery method, e.g., call list_ai_channels first]. If empty, all channels returned.'
• Document output schemas explicitly in tool docstrings and MCP registration. Example: 'Returns {channels: [string], values: {[channel_name]: number}, timestamp: ISO8601}'. This enables LLMs to plan chained calls.
• Add error response examples to docstrings. For get_ai_values: 'Raises ValueError if channel name invalid, LLM should call list_ai_channels to discover valid names. Raises RuntimeError if server unavailable, retry with exponential backoff.'
• Clarify optional parameter behavior. For channels param: 'If None or empty list, fetches all channels. Omitting this parameter is equivalent to passing empty list, call list_ai_channels first to see all available.'
• Group related tools and add discovery hints. In get_ai_values description: 'Call list_ai_channels first to see available channel names. For statistical confidence bounds, use get_ai_with_uncertainty instead.'
• Add acquisition_time constraints to get_ai_with_uncertainty: 'Acquisition time in seconds (float, range 0.1 - 10.0, default 1.0). Longer times improve precision; shorter times reduce latency.'