ESPHome device YAML authoring and esphome.io documentation navigation. Use when a request touches ESPHome firmware, device configuration, components (sensors, binary sensors, switches, lights, GPIO, I2C, SPI, UART, BLE, OTA, native API, MQTT), automations, lambdas, substitutions, packages, external components, or HA integration of ESPHome devices. Provides a structured sitemap of esphome.io so the agent fetches the correct page directly instead of crawling, plus practical working guidance for device YAML structure.
Fetch local Home Assistant config snapshots from ha-data/ via fetch-ha-data.sh, gated by a freshness check. Use when a request touches existing automations, scenes, scripts, configuration.yaml, or Core/Supervisor logs — none of which the MCP server exposes.
Output contract for Home Assistant implementation requests — the YAML / How This Works / Customization Ideas structure, runnable output only to .temp/, minimal valid YAML. Use when creating or editing automations, scripts, scenes, blueprints, or configuration.
Present Home Assistant live entity states with domain emojis, grouped by area or function. Use when answering current states, status checks, or monitoring queries from GetLiveContext output.
Docs-backed Home Assistant troubleshooting playbook — config check, logs, traces, Developer Tools, debug logging. Use when an automation, integration, or entity is failing, erroring, or misbehaving.
Structured sitemap of the official Home Assistant documentation (home-assistant.io/docs/) and the Companion App documentation (companion.home-assistant.io/docs/). Provides direct links to every major section and sub-page so agents can navigate to the correct docs page without crawling. Use when an agent needs to find the right HA documentation URL for automations, scripts, scenes, blueprints, templates, configuration, integrations, dashboards, energy, voice assistants, companion app notifications, sensors, location tracking, or troubleshooting.
Benchmark multiple AI model outputs from the ha-release-notes prompt against ground truth built from the user's live HA setup. Discovers model results, builds a release-specific scoring spec, evaluates each model on 7 weighted dimensions, and writes a ranked comparison. Use when user wants to compare, evaluate, score, or benchmark release note summaries from different models, or says "benchmark", "compare models", "score the results".