ChargingAtlas

Home / MCP

For ChatGPT & Claude

ChargingAtlas MCP server

ChargingAtlas runs a free, public Model Context Protocol (MCP) server, so AI assistants like ChatGPT and Claude can answer EV-charging questions straight from the U.S. Department of Energy's NREL data — every public station across all 50 states + DC. No app, no login.

Endpoint

https://mcp.chargingatlas.com

Streamable HTTP (JSON-RPC), no authentication — it's public DOE/NREL data. Read-only.

What you can ask

Tools

ToolWhat it does
search Find US cities by name — returns an id you pass to fetch. (ChatGPT/Deep-Research compatible.)
fetch The full EV-charging summary for a city id. (ChatGPT/Deep-Research compatible.)
get_city_charging A city's public charging: station count, Level 2 vs DC fast, networks, connectors, free vs paid, notable stations.
search_cities Find covered cities by name, optionally within a state.
list_state_cities Every covered city in a state, with station counts.
find_charging Filter/rank cities — by connector (J1772/CCS/NACS), network, DC-fast, or free charging.
nearest_stations The closest public stations to a latitude/longitude.
coverage_stats What's covered: states, cities, total stations, and the data date.

How to connect

ChatGPT

Settings → Connectors (developer mode) → add a custom connector and paste the endpoint above. No authentication. The search and fetch tools make it work with Deep Research.

Claude

Add it as a custom MCP server (Claude Desktop config or the claude.ai connectors settings) using the endpoint above — it needs no API key.

Prefer to just browse? Use the interactive map or find charging near you. Machines can also read our llms.txt.

Data: U.S. DOE / NREL Alternative Fuels Data Center. The MCP reports only what NREL records — it is not a live-availability map, so confirm a charger is working in the network's own app.