Stroomr Mcp

StroomR MCP

Helps AI assistants answer “when should I use electricity?” in one call — with live Nord Pool prices, local solar forecasts, and ready-to-use load-shift advice.

Built by StroomR. Specialized for European day-ahead energy markets and home energy timing — not a general weather or finance API.

---

What your agent can do

Job	Without this server	With StroomR MCP
EV charging advice	Agent guesses or asks user to check multiple apps	One tool call → optimal window, savings estimate, hours to avoid
Heat pump / battery timing	Manual price lookup + weather check	Combined price + solar overlap in structured output
“Is tomorrow sunny?” + “Are prices low?”	Two separate lookups, agent merges context	Single combined snapshot for load-shifting decisions

Outcome, not plumbing: your agent completes energy-timing tasks with fewer steps, pre-merged context, and structured fields (recommendedWindows, avoidWindows, estimatedSavingsVsAvgEur) instead of raw API dumps.

---

Best use cases

1. “When should I charge my EV tomorrow?” — cheapest consecutive window, estimated savings vs. average, evening peak to avoid.
2. “Optimal times for my heat pump today” — price-ranked hours with solar overlap for self-consumption.
3. “Will it be sunny tomorrow, and are electricity prices low?” — weather + Nord Pool snapshot for solar/load decisions.
4. “What are the cheapest hours to use power this week?” — day-ahead slots for today and tomorrow.
5. “How much sun will I get today?” — sun hours, sun windows, and rain periods for a given location.

Scope: Nord Pool day-ahead markets in Europe (NL, GER, DK1, SE3, etc.). Not for real-time trading, grid operator data, or non-European utilities.

---

Supported clients

Works with any MCP-compatible assistant, including:

• Cursor (local stdio — config included in this repo)
• Claude Desktop (add to claude_desktop_config.json)
• Other MCP hosts that support stdio servers (ChatGPT desktop integrations, custom agents, etc.)

Requires Node.js 20+. Runs locally on your machine — no StroomR account required to use the MCP.

---

Auth & safety

Runs	Locally via stdio (node dist/index.js)
Credentials	None — no API keys, no StroomR login
Network	Read-only outbound calls to Open-Meteo (weather) and Nord Pool public day-ahead API (prices)
Writes	None — does not control devices, meters, or your home
Data sent	Location (city or coordinates) and Nord Pool area — only what you pass in tool args or env config

Set a default location in env so the agent does not need to ask every time:

"env": {
  "STROOMR_CITY": "Amsterdam",
  "STROOMR_PRICE_AREA": "NL"
}

---

Examples

Example 1 — EV charging (primary use case)

User prompt:

When should I charge my EV tomorrow?

Agent action: get_load_shift_advice · { "device": "ev", "day": "tomorrow" }

Agent reply (from structured output):

Charge between 10:00–14:00 tomorrow. Estimated savings: €3.73 vs. average wholesale price for a 4 h session at 11 kW. Avoid 17:00–23:00 (evening peak). Overlaps with expected sunshine — good for solar self-consumption.

---

Example 2 — Heat pump today

User prompt:

Optimal times for heat pump today?

Agent action: get_load_shift_advice · { "device": "heat_pump", "day": "today" }

Agent reply:

Run intensive heating between 14:00–18:00 today (~€0.68 cheaper than average for 12 kWh). Avoid 19:00–23:00.

---

Example 3 — Weather + prices in one step

User prompt:

Is tomorrow good for cheap charging and solar?

Agent action: get_weather_and_prices · { "days": 2 }

Agent reply:

Tomorrow: ~15 h sunshine (07:00–21:00), day-ahead prices near zero mid-day. Strong day for shifting loads to midday.

---

Tools (5, purpose-built)

Tool	Agent job
get_load_shift_advice	Main tool. Optimal + avoid windows for EV, heat pump, or battery; savings estimate; solar overlap
get_weather_and_prices	Combined snapshot when the agent needs both sun and price context
get_energy_prices	Cheapest/expensive hours, current and tomorrow Nord Pool day-ahead slots
get_weather_forecast	Sun windows, rain periods, precipitation — when solar matters
resolve_location	Validate city/coordinates and resolve Nord Pool price area

Defaults for load-shift advice:

Device	device	Power	Duration
EV charger	ev	11 kW	4 h
Heat pump	heat_pump	3 kW	4 h
Home battery	battery	5 kW	2 h

Override with power_kw and duration_hours.

---

Install

From npm (after publish)

{
  "mcpServers": {
    "stroomr": {
      "command": "npx",
      "args": ["-y", "@nixman2/stroomr-mcp"],
      "env": {
        "STROOMR_CITY": "Amsterdam",
        "STROOMR_PRICE_AREA": "NL"
      }
    }
  }
}

From source

npm install
npm run build

Cursor (local dev)

.cursor/mcp.json:

{
  "mcpServers": {
    "stroomr": {
      "command": "node",
      "args": ["dist/index.js"],
      "env": {
        "STROOMR_CITY": "Amsterdam",
        "STROOMR_PRICE_AREA": "NL",
        "STROOMR_OUTPUT_FORMAT": "structured"
      }
    }
  }
}

Restart Cursor after changing MCP config.

Claude Desktop

{
  "mcpServers": {
    "stroomr": {
      "command": "node",
      "args": ["/absolute/path/to/stroomr-mcp/dist/index.js"],
      "env": {
        "STROOMR_CITY": "Amsterdam",
        "STROOMR_PRICE_AREA": "NL"
      }
    }
  }
}

---

Configuration

Variable	Description
STROOMR_CITY	Default city (geocoded)
STROOMR_LAT / STROOMR_LON	Default coordinates
STROOMR_PRICE_AREA	Nord Pool area (e.g. NL)
STROOMR_CURRENCY	Default EUR
STROOMR_TIMEZONE	Timezone override
STROOMR_OUTPUT_FORMAT	friendly (default) or structured

Priority: tool arguments → env vars → clear error (no silent default city).

Structured responses include a stroomr branding block and fields optimised for agent parsing (recommendedWindows, cheapestSlot, summary_nl).

---

Verify

npm run smoke-test

---

About StroomR

This MCP is a free demo of the load-shifting logic in StroomR — energy management that automatically times EV charging, heat pumps, and batteries against dynamic tariffs and solar production.

Want it to run 24/7 without asking an assistant? Join the free pilot →

---

Data sources & disclaimer

• Weather: Open-Meteo (CC BY 4.0)
• Prices: Nord Pool Group day-ahead via community-documented public API (unofficial; may change)

Load-shift advice uses wholesale day-ahead prices. Retail bills include taxes, grid fees, and supplier markups — savings estimates are indicative, not guaranteed.

Supported Nord Pool areas: DK1, DK2, FI, NO1–NO5, SE1–SE4, EE, LT, LV, AT, BE, FR, GER, NL, PL, BG, TEL

---

Publish to MCP Registry

Prepared for the official MCP Registry quickstart:

Step	Status
1. mcpName in package.json	✅ io.github.nixman2/stroomr-mcp
2. Publish to npm	⏳ Run locally (requires npm login)
3. Install mcp-publisher	✅
4. Create server.json	✅ Validated
5. mcp-publisher login github	—
6. mcp-publisher publish	—

Step 2 — publish to npm:

npm adduser          # if not logged in
npm publish --access public

Steps 5–6 — publish registry metadata:

mcp-publisher login github
mcp-publisher publish

Verify listing:

curl "https://registry.modelcontextprotocol.io/v0.1/servers?search=io.github.nixman2/stroomr-mcp"

Registry name: io.github.nixman2/stroomr-mcp · npm: @nixman2/stroomr-mcp

---

Submit to Cursor Directory

cursor.directory scans the repo root for Open Plugins components. This repo includes:

Path	Purpose
mcp.json	MCP server config (required for directory scan)
.cursor-plugin/plugin.json	Plugin manifest
skills/stroomr-energy/SKILL.md	Agent skill for energy-timing questions

After pushing to GitHub, re-scan at Submit a Plugin → Auto (GitHub) with https://github.com/nixman2/stroomr-mcp.

Note: mcp.json uses npx @nixman2/stroomr-mcp — npm publish must succeed first. Local dev still uses .cursor/mcp.json.