Privacy notice

In brief

• No application accounts, sign-in, address lookup or device geolocation.
• Only the parameters a selected tool needs are sent, and only to fixed NOAA and USGS services.
• The app does not persist or log request bodies, tool arguments or returned records. Relevant query parameters and results may be held in a bounded in-memory process cache.
• Your AI client and our websites are separate and follow their own policies.

What the connector receives

When a tool is called, the connector receives its inputs: for example a public research latitude and longitude, NOAA station IDs, date windows, requested units, datums and time zones, and response limits. It returns agency records with their source and retrieval details.

Please do not enter a personal address or the precise location of a home. Use public research points instead.

How a request flows

1. Your AI client, such as ChatGPT, sends a tool call with its arguments to the connector.
2. The connector sends only the query parameters needed for that operation to one of these fixed agency services: the NOAA CO-OPS Data API, the NOAA CO-OPS Metadata API, USGS EPQS, or the USGS Watershed Boundary Dataset service. Your conversation is not sent, and the connector does not pass along your client’s IP address as part of the query.
3. The agency responds, and the connector returns the records to your client with the original retrieval time.

Each agency processes requests under its own published policies, which may differ from one another. Source references: NOAA Tides & Currents, USGS EPQS, USGS Watershed Boundary Dataset.

Your AI client and our websites

ChatGPT or any other client you use processes your conversation and connection context under its provider’s policies, not this notice. These product and policy pages are hosted with the service. The GovBot website at govbot.ai is a separate site with its own hosting and policies. This notice does not claim that either collects nothing.

Caching and retention

No persistent payload storage

The app does not persist or log request bodies, tool arguments or returned record payloads.

Bounded memory cache

Recent results are held in process memory only: at most 64 entries and 32 MiB, with least-recently-used eviction.

Reuse windows

Ordinary queries may be reused for up to 10 minutes. NOAA station metadata and datums, and WBD service metadata, for up to 24 hours. Original retrieval timestamps and cache status are kept with each result.

Expiry is not timed deletion

These windows limit reuse. Expired entries are cleared lazily, so data may remain in memory until it is next accessed, evicted, or the process restarts.

No model training

The cache is not used to train models.

Operational logs

The app’s own logging is limited to a startup operational message; it does not log request or error payloads. The service is configured for Google Cloud Run. The inspected default operational-log sink routes to the project’s _Default bucket; its retention was verified as 30 days on 27 October 2026. Operational request logs may contain network or client IP addresses, request paths, user agents, times and response status. This describes the inspected default log configuration, not every provider, security or administrative audit record; those may follow separate policies. Source publishers and OpenAI also apply their own logging and retention practices.

Support email

Messages sent to hello@govbot.ai are kept until manually deleted; there is no fixed deletion schedule. You can email us to ask for help or to request deletion of your messages. We cannot promise automatic log deletion or removal of data from other providers’ systems.

Changes and contact

We will update this notice, including its date, as hosting details are verified or the service changes. Questions: hello@govbot.ai.