This is not a dashboard: `set_operation_mode`, `set_grid_export_limit` and `write_setting` change how the inverter behaves, so an assistant connected to it can alter what your house does with its own power. The upside is that questions like why the battery drained overnight can be answered against live sensor values instead of a vendor app, and the same session can act on the answer. Keep it on the LAN or behind a token and a TLS proxy.
A local server that talks to a GoodWe solar inverter over your own network, built on the `goodwe` Python library with sensor definitions and operation modes modelled on the Home Assistant integration. It reads every live sensor the inverter exposes and also writes: operation mode, grid export limit and battery depth-of-discharge are all settable.
- `get_runtime_data` returns all live sensor values, optionally filtered by kind — PV, AC, BAT, GRID, UPS or BMS
- `list_sensors` and `read_sensor` enumerate sensor IDs and read one at a time
- `get_settings_data`, `read_setting` and `write_setting` read and change any configurable setting
- `get_operation_mode` and `set_operation_mode` switch between general, eco, backup, off_grid, peak_shaving, eco_charge and eco_discharge
- `get_grid_export_limit` and `set_grid_export_limit` control export in watts, with 0 disabling it
- `get_battery_dod` and `set_battery_dod` set depth-of-discharge
- `connect_inverter`, `get_connection_status` and `get_device_info` handle the connection and report model, serial and firmware
Python 3.10+ and a GoodWe inverter reachable on the local network, on UDP port 8899 or Modbus/TCP port 502. Install with `uv pip install .` (package `goodwe-inverter-mcp`, version 0.1.0) and run `goodwe-mcp`. Set `GOODWE_HOST` to auto-connect at startup. Stdio is the default; `--transport sse`, `streamable-http` and `server` are available, and for those set `MCP_AUTH_TOKEN` — when it is empty, authentication is off and every HTTP endpoint is public. The server does not terminate TLS, so put a reverse proxy in front of anything not on localhost.
One command — uv pip install .