Security

Built for local-first trust boundaries.

Vaner is built around explicit scope, inspectable decisions, and predictable behavior in production workflows.

Local-first operation

Vaner is designed to run close to your codebase and workflow. You choose the workspace scope and can keep inference fully local when desired.

No content logging

Vaner does not log prompt/response content by default. Operational metadata and local traces are scoped for inspectability, not broad telemetry collection.

Explicit scope and exclusions

Context building is constrained by path allowlists and exclusion patterns. Sensitive files can be excluded proactively and redaction patterns can be enforced.

Inspectable context decisions

Every package decision can be inspected so you can see what context was selected and why. Vaner favors transparent preparation over hidden behavior.

Auto Focus boundaries

Auto Focus checks supported clients and Vaner integrations to decide whether Vaner should stay idle, standby, or prepare work for one active workspace. It does not read prompts, window titles, browser tabs, clipboard, terminal history, or arbitrary apps.

Preparation, not silent execution

Vaner prepares context. It does not silently run destructive actions in your environment. You stay in control of what gets executed and where.

Skill-aware scanning boundaries

When skill integration is enabled, Vaner scans SKILL.md files in repo-local roots (.cursor/skills, .claude/skills, skills). Paths inside the repo are treated as project-local. Global skill paths stay disabled unless you explicitly opt in with [intent].include_global_skills.

Current compliance posture

Vaner is early-stage software and does not currently claim SOC 2, HIPAA, or ISO certifications.