ScrubMyText sponge mascotScrubMyText
Agent infrastructure · Paid

ApproveMyAction
Put a human between your AI and the consequences.

Create a secure shareable approval link for a consequential action. The agent can poll the approval status and proceed only after the independent human decision.

1. Request “Refund customer $725”
2. Share the secure approval link with the reviewer
3. Human approves or rejects
4. Agent checks status before acting

Current MVP

  • Secure random approval links.
  • Pending requests expire after 24 hours by default; configurable up to 7 days.
  • Reviewer can approve, reject, or leave an optional note.
  • Agent can check status or cancel a pending request.
  • Approval details remain separate from the model's own reasoning.

What it does not do yet

The first version creates the secure approval link but does not automatically email or text it. The calling agent/application can send the returned URL through its normal communication channel. Email/SMS delivery can be added later.

Anyone who possesses a live approval link can make the decision, so it should be treated like a temporary secret and shared only with the intended reviewer.

Try ApproveMyAction

No approval created yet.
Result will appear here.

Creating, checking, or cancelling through the authenticated API each counts as one call. The human's approve/reject click does not.

REST endpoints

POST /v1/approvals/request
POST /v1/approvals/check
POST /v1/approvals/cancel

MCP tools

request_approval
check_approval
cancel_approval

Developer docs →

Read the human-approval guide →