Put a human decision between an AI agent and a consequential action.
Some actions should not depend on the model deciding that its own plan is acceptable. ApproveMyAction creates an external human-authorization step that the agent can verify before continuing.
The problem
An agent may be capable of issuing a refund, deleting data, changing permissions, or sending a high-impact message. For selected actions, the authorization decision should come from a person outside the model.
The pattern
1. request_approval(...) 2. Agent receives secure approval_url 3. Deliver link through your normal channel 4. Human approves or rejects 5. check_approval() 6. Continue only if approved
What ApproveMyAction provides
Secure approval URL
Each request gets a random secret link. Possession of the complete live link is the current authorization mechanism.
Clear decision state
The agent can check whether a request is pending, approved, rejected, expired, or cancelled.
Human note
The reviewer can optionally leave context with the approval or rejection.
No per-click usage
The authenticated request/check/cancel operations count as API calls. The human approve/reject click does not.
Use ApproveMyAction
Common questions
Does ApproveMyAction send the approval link?
No. The current product returns the secure URL to the calling application, which can deliver it through email, Slack, Teams, SMS, or another channel it already uses.
Does approval guarantee the action is safe?
No. It provides an external human authorization decision. Your application remains responsible for what the requested action actually does.