Get the current state of a Home Assistant entity (light, climate, sensor, switch, vacuum, media_player, etc.)

Control a light: turn on/off/toggle, set brightness (0-100%), color temperature, or RGB color

Control Tado thermostats: set target temperature or HVAC mode (heat/auto/off)

Activate a Hue scene in a room (wohnzimmer, flur, kuche, schlafzimmer, home)

Control the Roborock vacuum: start full clean, stop, return to dock, or get status

Read all environmental sensors: temperature, humidity, CO2 for all rooms plus outside temperature

Write a note to today's daily memory log on the OpenClaw server. Use for capturing important context, decisions, or things to remember.

Read today's memory log from the OpenClaw server

Search across daily memory files on the OpenClaw server for a keyword (default: last 14 days)

List all scheduled OpenClaw cron jobs

Trigger an OpenClaw cron job immediately by its ID

Get OpenClaw cron scheduler status and overview

Send a task or prompt to the OpenClaw AI agent via SSH. The agent has access to all installed plugins (trading, home automation, etc.) and multiple LLM backends. Use this to delegate complex tasks that OpenClaw already knows how to handle.

Check if the OpenClaw daemon is running on the server and get version info

List all plugins installed on the OpenClaw server

Send a prompt to Google Gemini via the local gemini CLI. Fast, direct LLM call with no OpenClaw overhead. Uses cached Google auth — no API key required.

Send a task to OpenAI Codex via the local codex CLI. Specializes in coding tasks, file operations, and technical analysis. Uses cached OpenAI auth — no API key required.

Send a prompt to Claude via the local Claude Code CLI. Use this when the MCP client is NOT Claude (e.g. Cursor, Windsurf, Zed) or for cross-checking results from other AI backends. Uses cached Anthropic auth.

Send a prompt to a local LLM (LM Studio, Ollama, llama.cpp, or any OpenAI-compatible server). Free, private, no API key needed. Best for simple tasks: classify, format, extract, rewrite, proofread. Set stream=true for token-by-token progress.

List all available elvatis-mcp tools with a routing guide. Optionally provide a task description to get a specific recommendation for which sub-agent (openclaw_run, gemini_run, codex_run, local_llm_run) or tool to use.

Analyze a complex prompt and split it into sub-tasks with agent assignments. Returns a structured plan showing which sub-agent (gemini, codex, openclaw, local LLM) handles each part, dependency ordering, and the actual prompts to send. Each subtask includes a suggested model that the user can override before execution. IMPORTANT: Always present the plan to the user for review before executing. Strategy: "auto" (default), "gemini", "local", or "heuristic".

Execute a prompt_split plan: runs subtasks in dependency order, dispatches to the correct sub-agent, passes results between dependent tasks, and enforces rate limits on cloud agents. Provide a "plan" from prompt_split, or just a "prompt" to generate and execute in one step. Use "overrides" to change agent/model/prompt per task or skip tasks. Set dry_run=true to preview.

Run a shell command on the configured remote Linux server (REMOTE_HOST). Use for deployments, log checks, service restarts, or any ad-hoc command on a remote machine.

Trigger deploy or rollback scripts on the OpenClaw server, or check the last deploy log. Scripts must exist at OPENCLAW_DEPLOY_SCRIPT_DIR (default: ~/deploy).

Manage Docker containers on the remote Linux server via SSH. Actions: list, logs, start, stop, restart, stats, exec. No Docker API or open port needed.

Manage systemd services on the remote Linux server via SSH. Actions: status, start, stop, restart, enable, disable, list.

Check health of all connected services at once: Home Assistant, OpenClaw (SSH), local LLM, Gemini CLI, Codex CLI. Returns a unified status overview with latency for each service.

List, load, or unload models on the local LLM server. "list" shows available models. "load"/"unload" switches models in LM Studio without opening the GUI (LM Studio only).

View recent logs from the OpenClaw server: gateway logs, agent execution logs, or system journal. Supports line count and keyword filtering.

List, trigger, enable, or disable Home Assistant automations. "list" shows all automations with their state and last trigger time.

Upload, download, or list files on the OpenClaw server via SSH. Supports text and binary files up to 10MB. "download" without local_path returns file content directly.

Send a notification via WhatsApp, Telegram, or the last-used channel. Uses OpenClaw message delivery.

Create a new cron job on the OpenClaw server. Supports cron expressions ("0 9 * * *"), intervals ("every 30m"), and one-shot ("at 2026-04-01T14:00" or "+20m"). Optionally deliver results via WhatsApp/Telegram.

Edit an existing cron job. Change its name, message, schedule, or model.

Delete a cron job by its ID.

Show recent execution history for cron jobs. Optionally filter by job ID.

Manage a local llama.cpp server: start with specific model, cache type (turbo2/turbo3/turbo4 for TurboQuant), GPU layers, and context size. Runs alongside LM Studio on a different port. Use "status" to check, "stop" to kill. Once started, use local_llm_run with the endpoint to query it.