AI Safe Mode
AI Safe Mode is a toggle that applies a set of local rules to your AI coding session — restricting which files and context assistants can read, warning on risky terminal commands, and requiring approval before sensitive context is shared. It turns ad-hoc caution into a consistent, reviewable policy.
Runs locally in your editor. Secret, PII, prompt-injection, and MCP scanning happen on your machine before anything reaches an AI model.
Enable or disable Safe Mode per session. When on, protected files and secrets are held back from AI context by default.
Choose terminal protection (manual, warn, or approval), file-size and workspace scan limits, and exclusion globs to fit your project.
Review, approve once, or deny AI context. Approvals are scoped to the session and can be cleared at any time.
“Show AI Safe Mode Rules” surfaces exactly which rules are active, so behavior is never a black box.
Enable Safe Mode
Turn it on from the command palette or the SoterAI panel at the start of a session.
Work as usual
When AI context includes protected files, secrets, or risky commands, Safe Mode warns or requires approval per your configuration.
Approve or deny in-flow
Grant session-scoped approval for context you trust, or deny it. All decisions are local.
Review and disable
Check the ledger of what was shared, then disable Safe Mode when the session ends.
No security tool is perfect. Here is what this feature does not claim to do, so you can layer defenses appropriately.
By configuration: sharing of protected files and detected secrets, and risky terminal commands. In approval mode it pauses those actions until you explicitly allow them.
No. Safe Mode rules are local. Team and enterprise policy sync is an optional, opt-in cloud feature.
Yes. Run “Show AI Safe Mode Rules” to view the active rule set at any time.
Install SoterAI IDE Guard and scan secrets, prompts, MCP tools, and terminal commands locally before they ever reach an AI model.
Install for VS Code