MCP server for controlling Valetudo robot vacuums via the Valetudo API
The Valetudo MCP server has weak definition quality overall. While all 7 tools are explicitly registered with names and descriptions present, there are significant gaps in schema quality, parameter documentation, and error handling. Tool names are reasonably clear and action-oriented (basic_control, get_robot_capabilities, etc.), but parameter descriptions are sparse. The zone_cleaning tool has a complex nested schema that is poorly documented. Many tools lack input validation guidance, and error messages are generic. Most parameters have basic descriptions but lack detail about valid values, ranges, or constraints. The server scores in the 'D' range (50-59 poor) due to incomplete schemas and minimal parameter documentation across tools.
Control basic robot functions (start, stop, pause, home)
Get the list of capabilities supported by the robot
Get basic robot information including manufacturer, model, and implementation
Get the current state of the robot including attributes and map
Locate the robot
Set the Valetudo log level
Start zone cleaning with specified zones and iterations
zone_cleaning has a deeply nested schema (points → pA/pB/pC/pD → x/y coordinates) that is poorly documented. Parameter descriptions don't explain what pA, pB, pC, pD represent (corners of a zone) or coordinate system units.
locate tool has only a 5-word description ('Locate the robot') with no explanation of what 'locate' means (e.g., trigger audible beep, return GPS coordinates, etc.) or what output is returned. Ambiguous.
basic_control's 'action' parameter description is minimal ('Action to perform (start, stop, pause, home)'). No explanation of when each action is valid, what state transitions are allowed, or what happens if you call 'stop' twice.
set_log_level's 'level' parameter description does not explain what effect each log level has, whether changes persist across restarts, or if the setting affects all subsystems or specific components.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 42 | - | v1 |
Output schemas are not documented for any tool. Clients cannot know what fields to expect in responses or plan downstream operations. E.g., does get_robot_state return 'battery_level' or 'batteryLevel' or 'battery.percentage'?
Error handling is generic. All tools return errors like 'failed to get robot state: <error>' with no recovery guidance. LLM cannot distinguish between retryable errors (network timeout) and unrecoverable ones (robot offline, invalid parameter).
No input validation hints in parameter descriptions. E.g., zone_cleaning doesn't explain coordinate ranges (0-1000? 0-100? meters or pixels?), and basic_control doesn't validate that 'action' is one of the listed values before calling the API.
set_log_level accepts free-form strings for 'level' parameter. Should declare as an enum: ["trace", "debug", "info", "warn", "error"] to prevent hallucinated values like 'verbose' or 'silent'.
basic_control tool description does not clearly state it modifies robot state (WRITE operation). LLM might not realize calling 'stop' while robot is docking could interrupt charging.
zone_cleaning description does not explain if iterations > 1 causes overlapping passes, if zones are cleaned in order, or what happens if zone coordinates are invalid. Parameter 'zones' is complex but underdocumented.