Privacy policy

Data we collect

The server asks for no sign-in. It receives a network address and the arguments sent with each tool call. To answer a request it receives the tool name and arguments your AI assistant sends, such as a search phrase, a table number or a city name. It does not automatically receive your full conversation, but arguments can contain text from it. The data it returns comes from public sources.

How the data is used and stored

• Tool arguments are used to run the request. Tool failure logs omit argument values and exception text. HTTP access logs can include the network address, method, path and status of a request.
• A count of tool runs is kept in memory by tool name, day and success or failure, and is public at /stats. It holds no arguments, no client addresses and nothing that identifies a person. It resets when the server restarts.
• A client's network address is held in memory to apply the rate limit (60 requests a minute). Entries with no requests in the last minute are removed when another client makes a request; they can remain longer while the server is idle.
• Recent responses from the public sources are cached in memory to save time. The cache can include search phrases and filters in its keys or responses. Entries expire or are removed as space is needed, and the in-memory cache disappears when the server stops. Public source files can also be cached on temporary disk until removed or cleared by the host.

Retention

The in-memory items above last until expiry, cleanup or the end of the server process. Public source files can remain on temporary disk. HTTP access logs are retained under the host's practices. Files that a tool returns, such as an Excel workbook, are sent to you in the response and are not kept.

Who else sees a request

• The publisher of the data. To run a tool, the server sends a request to the publisher's service, such as Statistics Canada, the Bank of Canada or a city's open-data portal. The publisher sees that request and the server's address, not yours, under its own terms.
• The host. The endpoint runs on Render, which keeps its own request logs under its privacy policy and terms.
• Your AI assistant. Claude or any other client handles your conversation under its own privacy policy. MapleStats receives only what the client includes in tool calls.

MapleStats does not sell data, share it with advertisers, run analytics or tracking services, or use requests to train models.

Keeping everything on your machine

MapleStats is open source. If you run it locally, nothing reaches the hosted server, and the only outside requests are those to the publishers.

Children

The service is not directed at children. The request handling described above applies to all users.

Changes and contact

Changes to this policy are published on this page with a new date, and the full history is in the repository. For questions, open an issue; for a security problem, use the private vulnerability report. The maintainer is Daniel Sánchez Pazmiño.