Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Meter MCP Server classify today's 96 readings for meter M-123 as active, standby, off, or unsure"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Frozen meter-day classifier behind MCP. 90.0% eval vs 86.7% rules floor. Offline, small, and fast.
One sklearn logreg (C=1.0, 26 evidence features) exposed as 3 MCP tools:
classify_day - 96 x 15-min kWh plus meter reference returns label plus reason.
classify_batch - up to 31 days.
model_info - version, numbers, limits.
Labels: active, standby, off, unsure. unsure is a first-class abstention, never a forced call.
Status: Step 0 scaffold. classify_* are stubs until Step 1 vendors the frozen model.
Related MCP server: WorkloadTruth MCP Server
pip install -r requirements.txt
export PYTHONPATH="$PWD/src"
python3 -m pytest -q
bash reproduce.shexport PYTHONPATH="$PWD/src"
python3 -m meter_mcp.server_stdioMCP_API_KEY=test python3 -m meter_mcp.server_http --port 8000
curl -s localhost:8000/healthz
docker build -t meter-mcp .
docker run -p 8000:8000 -e MCP_API_KEY=test meter-mcpSealed 30 day eval from the sibling repo: 90.0% (27/30), decline 36.7% at precision 0.727, validity 1.0, p50 0.77 ms per day, zero confident errors. Sibling: https://github.com/TMFNK/meter-day-classifier
What the numbers do not show: day labels are the benchmark, not a customer savings claim. Segment triage plus a real pilot come later.
See 2026-09-29-implementation-plan in the vault (20_Projects/40_mbitai-Meter-MCP).
Apache-2.0. See LICENSE and NOTICE. MbitAI, Munich.
This server cannot be deployed
Read-only electricity, gas, and weather data with structured provenance and units.
Crypto regime classifier for trading agents. 100% bounded regime accuracy. 50 free calls/day.
Classification service to seamlessly retrieve SOC codes for job titles and descriptions.
Structured analysis API and remote MCP tool for text, JSON records and numeric series.
