withHuman

withHuman is an open-source approval layer for AI agents. It intercepts agent tool calls before execution, automatically approving or denying routine calls based on rules and routing risky actions to the right person on your team. It works across tools like Claude Code, Codex, and agents built with LangGraph or Pydantic AI through an open protocol (AAP), letting companies apply one set of rules across all their agents. It is available as open source with a hosted version.