Harden (Agentic Integrity Foundation)
A local guardrail for AI coding agents: it intercepts every agent action before execution, blocks dangerous commands and masks sensitive data.
Description
Harden (Agentic Integrity Foundation) is a local security tool for AI coding agents. It intercepts tool calls and commands before they run, judges whether they are safe, blocks malicious operations and masks sensitive data, keeping an audit log.
Key features
- Intercepts and evaluates agent actions before execution
- Blocks dangerous and malicious commands
- Masks confidential data
- A local per-session audit log
- Compatible with Claude Code, Codex, Cursor and more
Pricing and access
The tool is free and runs locally. It is built by a team with experience at Google DeepMind, Amazon Alexa, Zscaler and CrowdStrike.
