How Board Badger uses AI.

Board Badger is built to be driven by an AI assistant, which makes it worth being exact about where a model is involved and where one is not.

There is no model inside Board Badger

Nothing in the running service sends a board to a language model. The command line on a board reads a sentence with a fixed set of patterns and lands it in one of the same fourteen tools an agent would call. It refuses rather than guessing when it cannot tell what was meant.

Your assistant is yours

Board Badger is an MCP server, so an assistant an account holder already uses, Claude included, can read and update their boards on their behalf. That assistant is the account holder's own relationship, not the Operator's. What it does with what it reads is governed by whoever provides it, and a board is only ever reached with an OAuth token issued to that account after a consent screen.

What an agent can and cannot do

The fourteen tools follow the same rules the web application follows, and each is scoped to a board the account owns. An agent can put a task on, change its title or description, move one between the three states, tick one off, comment, and invite somebody at the client. Files are added on the web page, not through a tool. It cannot reach a board on another account, and no tool deletes a board.

What a client sees

An agent writing on a board is writing where the client can read it. A comment added by an assistant on an account holder's behalf appears under that account holder's name, because it was theirs to make. Work that should not be seen belongs on a board nobody was invited to.

Training

Boards are not used to train any model, and are neither sold nor shared. The Privacy notice sets out what happens to personal data generally.