For developers & AI agents

Query TapWaterMap from your tools

TapWaterMap runs a free, public Model Context Protocol (MCP) server so AI assistants — ChatGPT, Claude, or any MCP client — can query US tap-water quality directly from the EPA's own data. No signup, no API key.

The endpoint

Streamable-HTTP MCP, public and read-only:

https://mcp.tapwatermap.com/mcp

Connect it

• ChatGPT: Settings → Connectors (Developer Mode) → Add → paste the URL above → no authentication.
• Claude: Settings → Connectors → Add custom connector → paste the URL → no authentication.
• Any MCP client: point it at the endpoint; tools are discoverable via tools/list.

Tools

• search / fetch — find a city, then fetch its full record (ChatGPT-compatible).
• get_city_water — the systems serving a city and what the EPA has on record.
• get_water_system — a single water system by PWSID.
• search_cities / list_state_cities — discovery by name or state.
• find_cities — filter/rank across the dataset (by state, contaminant, health-based, population, open violations).
• compare_cities — compare several cities side by side.
• explain_contaminant — what a contaminant is, plus whether the EPA classifies it as health-based.
• coverage_stats — what's covered (states, cities, records, data quarter).

Data & the rules of the road

• Source: U.S. EPA Envirofacts SDWIS — the same data behind every page here. ~19,000 cities, all 50 states + DC, ~1.5M violation records.
• Every response reports what the EPA recorded — never a safety verdict. The absence of a violation is not a guarantee of safety. Please preserve that framing when you cite it.
• Read-only and public. Be reasonable with request volume.

Machine-readable summary for agents: /llms.txt. How the data is built: Methodology & Sources.