withHuman

withHuman lets a company put the riskiest actions of its AI agents behind human approval. When an agent decides to make a consequential tool call, such as refunding a customer through Stripe or paying a supplier's invoice, withHuman intercepts the call before it reaches the payment or business system. An approval pipeline then decides: rules can approve some calls automatically, for example refunds under a set amount, and send the rest to a person. Calls are routed to the right team and to whoever is on call there, and that reviewer is notified on the channel their own rules choose, including Slack, Discord, email, text message, WhatsApp and mobile apps for iPhone and Android. The reviewer sees the tool call and the context of the request before approving or denying it; an approved call goes through and the agent carries on. The maker describes the product as framework-agnostic, open source and built on open standards. Its site presents it as launching on 15 October 2026 and states no prices and no company location.

Not yet independently verified: The launch date rests on the Product Hunt daily leaderboard of 15 October 2026 and the maker's own "launching today" notice. Capabilities come from the maker's own site and were not independently checked, and it states no price. We will update this when it can be confirmed, and remove this note.