Set policies: allow, warn, block

The Policies page controls how Northbeams responds to AI use. Rules apply across all four surfaces in real time.

The three actions

• Allow: the tool runs, and the activity is recorded.
• Warn: the user sees a nudge and can proceed. You get the signal without blocking work.
• Block: the action is stopped.

How rules are set

• By category. Set a default action for each category of tool, so a new tool in that category is governed the moment it appears.
• Per tool. Override the category rule for an individual tool when the broad rule is too strict or too loose.
• Network Control. Add DNS-layer rules that catch shadow-AI domains before a tool even loads.

Govern, do not block by default. Most teams start with warn on unknown tools and block only on the high-risk ones, then tighten from there.

Next: govern the tools your coding agents call with the MCP Gateway.