MCP server for controlling and querying Hubitat Elevation home automation hub devices
This server has 6 tools with visible definitions in src/hubitat_mcp/server.py. All tools have basic descriptions (10-80 chars) and input schemas. However, critical gaps undermine quality: (1) Tool names lack clarity - 'list_devices' and 'device_details' are generic discovery tools but lack the rich return-type documentation agents need; 'device_history', 'device_capabilities', 'device_commands' are all variations on 'get device property X' without explaining when to use each; (2) Parameter descriptions are minimal (e.g., device_id: 'The ID of the device to get details for' - no format hints, no guidance on how to obtain the ID); (3) Output schemas are completely undocumented - tools return JSON via json.dumps() but callers have no specification of returned fields, structure, or pagination; (4) Error handling is absent - requests.get(url) with no error checking means HTTP failures (401, 404, 500) return raw status codes or empty JSON without guidance; (5) No validation on control_device command parameter - accepts free-form strings like 'setLevel/50' or 'setCode/3,4321,Guest' with no enum, pattern, or format constraints. The descriptions are functional but fall short of the LLM-optimized 50-200 char baseline for A+ tools. Baseline comparison: average tool description in production (194 chars) vs actual (avg ~45 chars). This server is below median community quality.
Command the Hubitat Devices
Get Capabilities for a Specific Device
Check Commands for a Specific Device
Check Specific Device Details
Check Event History for a Specific Device
List the Hubitat Devices
No output schemas documented. All tools return json.dumps() of API responses, but callers cannot determine what fields, types, or structure to expect. This violates the critical pattern that 'LLMs need to know what fields to expect so they can plan downstream tool calls.'
Parameter descriptions lack actionable detail. Example: 'device_id' parameter across all tools is described as 'The ID of the device to...' with no format hints, example, or guidance on how to obtain the ID. Baseline: production tools average 72 chars per param description; these are ~30-45 chars and lack specificity.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 8 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 42 | - | v1 |
control_device accepts free-form command strings (e.g., 'setLevel/50', 'setCode/3,4321,Guest') with no enum constraint, regex pattern, or validation. This invites hallucinated command syntax and fails silently on typos. Should either enum valid commands per device type or validate early with actionable errors.
No error handling guidance. All tools call requests.get() without try/except and return raw JSON on success or HTTP error codes on failure. LLM receives no actionable direction: e.g., on 404 (device not found), should it retry, ask for a different device_id, or list available devices? Critical missing recovery guidance per pattern:recovery-guide.
Similar tool names invite LLM confusion. 'device_history', 'device_capabilities', 'device_commands' are all 'get property X for a device' but lack distinguishing language about WHEN to call each. Should add use-case hints: 'device_history' = past events; 'device_capabilities' = what the device can do; 'device_commands' = how to control it.
No API response filtering or pagination. All tools dump raw Hubitat API responses via json.dumps(). If list_devices returns 100+ devices, the response bloats the context. Should cap results (e.g., top 20), offer pagination params, and strip irrelevant fields.
tool descriptions are generic and under the production baseline. Baseline: 194 chars avg for production tools. Actual: 'List the Hubitat Devices' (24 chars), 'Check Specific Device Details' (28 chars). These are functional but do not answer 'WHEN to use this' or 'WHAT does it return'. Per LLM-optimized guidance, descriptions should be 50-200 chars and include use-case hints.