withHuman

withHuman is the approval layer for AI agents. When an agent wants to do something consequential, a refund, a deploy, a customer email, a change to production, it asks first. This app is where you answer.

Every request shows exactly what the agent wants to call and with what arguments, the agent's own reasoning, and the trail of what has already happened. You approve or deny with one tap, add a note for the audit log, and the agent continues. Nothing runs until a person says so.

Made for people who are on the hook for what their agents do: engineers reviewing deploys, finance teams approving refunds and plan changes, support leads signing off on customer communication, security teams gating credential and access changes.

What you get

- A review queue that shows only what needs a decision, with resolved and expired requests one tap away.
- The full request: tool, arguments, reasoning, policy scope, and audit trail, rendered so machine content reads as machine content.
- Approve or deny with an optional note. Decisions are recorded with who, when, and from which channel.
- Claim a request so other reviewers know you are on it.
- Push notifications the moment a request needs you.
- Sign in with your work email and password, or with your company's single sign-on.
- Light and dark appearance, following the system or your choice.

withHuman is for organizations that use the withHuman platform. Your organization connects its agents and invites you; there is no separate sign-up in the app. The platform is built on an open protocol, and the self-hosted edition is open source.