Developers using AI coding assistants may unintentionally expose proprietary code through autocomplete or chat features. A command-line tool could toggle these features off during work on sensitive repos.
The tool would hook into IDE APIs to disable cloud-based suggestions when it detects access to directories marked as confidential.
Sell to enterprise security teams as part of compliance toolkits, with annual licensing.
MVP: A simple script that kills local AI coding agent processes when specific git remotes are active.
Risk: Developers may bypass the tool if it disrupts workflows too aggressively.