A Model Context Protocol (MCP) server for residential heat pump sizing, cost estimation, and cold-climate performance verification. Use with AI assistants like Claude Desktop, Claude Code, Cursor, and other MCP clients.
Works out-of-the-box with bundled data - no API keys required!
The server runs directly via uvx - no installation required. Your MCP client will handle this automatically.
Choose your preferred AI assistant:
Claude Desktop (Anthropic)
Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"heatpump-calculator": {
"command": "uvx",
"args": ["--refresh", "--from", "git+https://github.com/subspace-lab/heatpump-mcp-server.git", "heatpump-mcp-server"]
}
}
}Restart Claude Desktop.
Claude Code (VS Code Extension)
Add to .claude/mcp.json in your workspace:
{
"mcpServers": {
"heatpump-calculator": {
"command": "uvx",
"args": ["--refresh", "--from", "git+https://github.com/subspace-lab/heatpump-mcp-server.git", "heatpump-mcp-server"]
}
}
}Restart VS Code.
Cursor (AI Code Editor)
Add to Cursor's MCP settings (Settings > MCP):
{
"mcpServers": {
"heatpump-calculator": {
"command": "uvx",
"args": ["--refresh", "--from", "git+https://github.com/subspace-lab/heatpump-mcp-server.git", "heatpump-mcp-server"]
}
}
}Restart Cursor.
Other MCP Clients
Any MCP-compatible client can use:
{
"mcpServers": {
"heatpump-calculator": {
"command": "uvx",
"args": ["--refresh", "--from", "git+https://github.com/subspace-lab/heatpump-mcp-server.git", "heatpump-mcp-server"]
}
}
}That's it! The server includes:
- 81 heat pump models from major manufacturers
- 2024 state-average electricity rates for all US states
- TMY3 weather station data for climate zones
No API keys needed to get started.
calculate_heat_pump_sizing: Single-zone BTU sizing with humidity considerationscalculate_multi_zone_sizing: Floor-by-floor load calculations for complex homesestimate_energy_costs: Bill comparison and 10-year payback analysischeck_cold_climate_performance: Verify capacity at design temperatureget_electricity_rate: Fetch current electricity rates by ZIP codelist_heat_pump_models: Browse 81 heat pump models with specs
design-temps/{zip_code}: Climate data and design temperaturesheat-pump-models: Complete model database with BTU, HSPF2, pricesclimate-zones: ASHRAE climate zone reference
size-heat-pump: Step-by-step sizing guidanceanalyze-costs: Cost comparison workflowverify-cold-climate: Cold climate suitability check
User: I need help sizing a heat pump for my 2000 sq ft home built in 1995 in ZIP 02138.
AI: [Uses calculate_heat_pump_sizing tool]
Based on your location (Cambridge, MA - Climate Zone 5A) and home characteristics:
- Required BTU: 80,000 BTU
- Recommended range: 72,000 - 88,000 BTU
- Design temperature: 6Β°F
...
User: What would a Mitsubishi MXZ-3C30NA cost to operate vs my gas furnace?
AI: [Uses estimate_energy_costs tool]
Annual cost comparison:
- Heat pump: $1,850/year (using bundled state average rate)
- Gas furnace: $2,400/year
- Annual savings: $550
- Payback period: 8.2 years
...
User: Will a Fujitsu AOU24RLXFZ work in Minneapolis?
AI: [Uses check_cold_climate_performance tool]
Cold climate analysis:
- Design temp: -13Β°F
- Heat pump capacity at design: 18,000 BTU
- Your heating load: 75,000 BTU
- Coverage: 24% (Inadequate)
- Recommendation: You'll need substantial backup heat...
For more accurate electricity rates, you can optionally provide an EIA API key:
-
Get a free EIA API key: https://www.eia.gov/opendata/register.php
-
Add to your MCP client config:
{
"mcpServers": {
"heatpump-calculator": {
"command": "uvx",
"args": ["--refresh", "--from", "git+https://github.com/subspace-lab/heatpump-mcp-server.git", "heatpump-mcp-server"],
"env": {
"EIA_API_KEY": "your_eia_api_key_here"
}
}
}
}Or create a .env file in your working directory:
# Optional: For live electricity rate lookups
EIA_API_KEY=your_eia_api_key_hereNote: Without an API key, the server uses 2024 state-average electricity rates, which are generally accurate for cost estimates.
git clone https://github.com/subspace-lab/heatpump-mcp-server.git
cd heatpump-mcp-server
uv pip install -e .If you prefer to install the package locally instead of using uvx:
# Install from GitHub
uv pip install git+https://github.com/subspace-lab/heatpump-mcp-server.git
# Then configure your MCP client to use the local installation:
{
"mcpServers": {
"heatpump-calculator": {
"command": "heatpump-mcp-server"
}
}
}Note: Using uvx with --refresh is recommended for most users as it automatically updates to the latest version.
Built on FastMCP for easy MCP server development.
- Heat Pump Models: Bundled database of 81 models (Mitsubishi, Fujitsu, Daikin, LG, etc.)
- Climate Data: TMY3 weather station database with design temperatures
- Electricity Rates: Bundled 2024 state averages (EIA API optional for live data)
- Sizing: Climate-zone specific BTU/sqft coefficients based on building age and insulation
- Costs: Monthly degree-day analysis with heat pump COP curves
- Cold Climate: Manufacturer capacity curves with temperature derating
# Clone repo
git clone https://github.com/subspace-lab/heatpump-mcp-server.git
cd heatpump-mcp-server
# Install with dev dependencies
uv pip install -e ".[dev]"
# Run tests
pytest
# Lint code
ruff check .heatpump_mcp_server/
βββ src/heatpump_mcp_server/
β βββ server.py # Main MCP server
β βββ tools.py # Calculator tools
β βββ resources.py # Data resources
β βββ prompts.py # Guided prompts
β βββ config.py # Configuration
β βββ models/ # Pydantic models
β βββ services/ # Business logic
β βββ quick_sizer_service.py
β βββ bill_estimator_service.py
β βββ cold_climate_service.py
β βββ ...
βββ data/ # Bundled data files
β βββ hpmodels.json # 81 heat pump models
β βββ eeweather_stations.json # Weather data
βββ tests/
βββ pyproject.toml
βββ README.md
Contributions welcome! Areas for improvement:
- Additional heat pump models
- More weather stations for better coverage
- Manual J load calculation support
- International climate zone support
MIT License - see LICENSE file for details.