How healthy are
your AI agents?
Preamp scans your Claude Code, Cursor, Windsurf, and other AI agent configs for security risks, model waste, and misconfigurations. Free, instant, private.
What your audit looks like
Agent Audit Report
Scanned 4 agents · 12 configs · 3 devices
April 7, 2026
Four categories, zero blind spots
Exposed secrets & injection risks
Detects API keys, tokens, and credentials in config files. Flags prompt injection patterns and suspicious content in custom skills.
Right model, right cost
Analyzes each agent's purpose and recommends Haiku, Sonnet, or Opus. Shows estimated cost savings when you're overpaying for simple tasks.
Valid configs, no dead weight
Checks file format validity, identifies stale configs that haven't changed in 90+ days, flags bloated files, and spots duplicated settings across tools.
Industry-standard patterns
Checks CLAUDE.md structure, skill organization, MCP server configuration, and memory management against recommended patterns.
Get your audit in 60 seconds
Install Preamp
One command. Works on macOS and Linux.
Run the audit
preamp audit scans every agent config it can find.
View your report
Results open in your browser with full details and fix suggestions.
Running a team?
See audit results across every developer's agents. Enforce security policies, standardize model usage, and catch config drift before it causes problems.
Learn about TeamsGet your audit in 60 seconds.
Install Preamp, run the scan, see your results. Free forever.
Install & Run Audit