Verifiable execution for autonomous machines.
9192 lets agents run paid, verifiable machine calls and receive receipts they can check later.
Useful where machines need proof.
9192 fits agent execution, receipt verification, timestamp proofs, machine event logs and bounded paid compute.
What is this for?
Use 9192 when a machine action should end with evidence another client can check later.
One useful path, end to end.
The first run is free and limited: create a sandbox quote, accept it, execute a tiny call, receive a receipt, then verify it.
Read status, OpenAPI, agent metadata and pricebook.
Ask for measured work before execution.
Move from preview to an accepted quote.
Run the sandbox or funded service path.
Receive execution evidence.
Validate the receipt through the public API.
Interactive sandbox.
This calls the public sandbox routes. No funding, no paid execution, no account settlement.
Create a sandbox quote to begin.
Operational surface.
Public checks stay visible because machines should not need guesswork before integration.
Capabilities, not decoration.
Every capability is classified by use: public, authenticated, paid or administrative.
Integration paths.
Use the path that fits the client: browser-safe sandbox, HTTP API, native TLS CLI or remote MCP metadata.
Trust is a workflow.
Release artifacts are signed, checksummed and linked from machine-readable discovery. Receipts are verified after execution.