Code analysis and local findings
Free checks and supported Pro hooks evaluate code locally. Free does not require an account or source upload. Local findings and repair verification stay in your workflow.
Security & data handling
How Fettl handles your code.
Your source does not need to leave your machine for Fettl to inspect it. Local analysis and account services have distinct roles.
Review the source, access, and repair boundaries before introducing Fettl to your workflow.
Data handling
Local analysis and connected services.
Identity and Pro access
Pro login uses the browser and account service to issue signed access. The CLI stores credentials through the operating system’s secure credential store. Valid signed local access does not require an account-service call for every agent event.
Deterministic repairs
Apply. Check. Keep or restore.
Explore what happens when a repair passes, fails, or meets changed code.
normalize.tsRepair example · CQ003
// TODO: implement normalizeLabel export function normalizeLabel(label: string) {
return label.trim().toLowerCase();
}fettl fix --filter CQ003 --dry-runPlanned
Remove the stale TODO.
The function already works. Only its outdated comment needs to go.
If recovery needs attention
If Fettl cannot confirm that the original files were restored, it keeps backups and reports what needs recovery.
Fettl fixes eligible defects. Architecture and product decisions stay with you.
Agent hooks
Checks at the moments that matter.
Six native events across four agents. Pro connects them to your local checks.
Claude CodePreToolUsePostToolUseStop
PreToolUse- Checks a proposed shell command or edit. Can block an unsafe action before it runs.
PostToolUse- Checks the result of a shell command or edit and returns findings to the agent.
Stop- Checks completion evidence when the assistant finishes.
Codex CLIPostToolUse
PostToolUse- Checks completed tool work and returns findings for the agent’s next action.
OpenCodefile.edited
file.edited- Checks the file that was just edited and returns findings to the agent.
Hermespost_tool_call
post_tool_call- Checks code after write_file or patch completes and returns findings to the agent.
Post-tool checks act on completed work. Missing or incomplete analysis is reported explicitly.
Planned: richer agent steering
User-configured model providers, confirmed rules from your feedback, questions when intent is unclear, and synced profiles.
Security questions
Discuss your security requirements.
Ask about data handling, access, or repairs.
Ask a security question