How do you let an AI assistant query your data safely?

Connect it through the MCP server, which lets assistants such as Claude, Cursor and Codex read your data using your own permissions. It is read-only by default, and any change the assistant drafts becomes a proposal you approve in the Portal.

Video: Connect your AI to mWater on YouTube

Why this matters

The appeal of asking questions in plain language is obvious, and so is the risk: an assistant holding a token that quietly rewrites production records. The design answers that directly. The assistant sees only what your own account can see, nothing is written without human approval, and the token can be revoked at any moment. That is what makes it usable against real monitoring data rather than a copy of it.

How to do it

  1. Generate an MCP token on your account page in the Portal.
  2. Add the token to your AI tool, such as Claude Desktop, Cursor or Codex.
  3. Ask questions about your sites, forms and records in plain language.
  4. Analyse monitoring data without writing queries or exporting it first.
  5. Review and approve any proposed change in the Portal before it takes effect.
  6. Revoke the token whenever you need to cut off access.

Go deeper