Qorami sits between your AI agent and the send button, giving you a human checkpoint before emails go out. The MCP server exposes verify_email and check_action_status tools that return one of three verdicts: send it now, wait for human approval, or block it entirely. You get prompt injection detection out of the box, plus policy profiles to enforce different rules for different contexts. If the decision is request_human_confirmation, poll check_action_status until a human reviews it in the Qorami dashboard and flips it to send. Reach for this when you're shipping agentic workflows that compose emails but you're not ready to let them fire unsupervised.
claude mcp add --transport stdio io.github.loicfontaine-max-qorami uvx qorami