Full AI debugging and maintenance access to your Home Assistant via MCP
Hamster MCP defines 6 tools with complete input schemas and descriptions. Tool names follow verb_noun convention (search, explain, call, schema, list_resources, read_resource). Descriptions are present and contextual (avg ~120 chars), exceeding the 10-char minimum. However, parameter descriptions lack specificity about constraints, formats, and error recovery. The 'call' tool accepts an open-ended 'arguments' object with minimal guidance on valid keys. Output schemas are not documented, LLMs cannot predict response structure for planning. Error handling is absent from descriptions. Overall: solid foundation with gaps in parameter validation guidance and output documentation.
Execute a command. Path format: group/command. Arguments are command-specific.
Get detailed description of a command. Path format: group/command (e.g., 'services/light.turn_on', 'hass/get_states').
List available resource documents with practical guidance for working with Home Assistant entities, services, and configurations. Returns URIs that can be read with read_resource.
Read a specific resource document by URI. Use list_resources to discover available URIs.
Get schema/type information for a command or type. For services, use 'services/selector/TYPE' (e.g., 'services/selector/duration'). For commands, use the command path.
Search for commands across all groups. Use path_filter to narrow scope (e.g., 'services', 'services/light', 'hass/config').
The 'call' tool accepts an open-ended 'arguments' object with no schema constraints or guidance on valid keys. LLMs cannot predict what arguments a given command accepts without calling 'explain' or 'schema' first, forcing multi-step discovery.
No output schemas documented for any tool. LLMs cannot plan downstream operations or extract required fields (e.g., does 'search' return command paths, descriptions, or both?). This violates the response-shaper pattern.
Parameter descriptions lack format/constraint guidance. 'path' parameters state 'group/command format' but do not specify valid group names (services, hass, supervisor?), character restrictions, or error cases. LLMs must guess valid formats.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 69 | 2026-07-28+ | v2 |
No error handling guidance in tool descriptions. If a user calls 'explain' with an invalid path, what error is returned? How should the LLM recover? Descriptions do not answer these questions.
The 'call' tool modifies state (WRITE risk) but its description does not warn of irreversible consequences or suggest confirmation/dry-run patterns. Agents cannot distinguish safe reads from destructive writes.