Paste or drop a Testimony Record below and this page will report the level it reaches and every check behind that level. The validator runs in the page. Nothing is uploaded, and nothing is stored, so a record about somebody’s customers does not leave the machine it is read on.

It is the same validator the specification publishes, ported to JavaScript and checked against the Python reference on a corpus of records covering every level and every failure mode. If the two ever disagree the build fails, because two validators that disagree are worse than one.

Do not have a Testimony Record yet? Review reads the logs a system already writes, in whatever shape it writes them, and reports what they establish before you produce anything in this format.

Any system can emit one and the format requires nothing of ours. Free adapters exist today for LangGraph, CrewAI, the OpenAI Agents SDK, Pydantic AI and AutoGen. For a LangGraph graph that already pauses for a person, it is an adapter and about five minutes:

pip install testimony-langgraph

The node names the action so the operator’s own risk table can classify it, and the identity of whoever approves is taken from the authentication layer rather than from anything the model produced. The adapter will not invent an approver, because there is nothing in LangGraph for it to find, so a graph resumed the ordinary way produces a record that says so rather than implying a person was involved.

Not sure which of the five applies to your framework? instrument quickstart --framework <name> names the right adapter, prints its install line and wiring snippet verbatim from that adapter’s own README, and chains straight into a check on the record it just produced. The library still only measures; the emitter, which stays free and separate, is what writes.

The same validator ships with each adapter as testimony-validate, so a record can be checked at a terminal as well as here.

A level is a property of the record, not a judgement about the system that wrote it. TR-3 says an approval named a person and took that name from an authentication layer. It does not say the person read anything, and no format can.

Below TR-4 nothing here is tamper evidence. Non-decreasing write times show a record is consistent with having been appended to, not that it was. And a record that declares it takes no actions is believed, which is why the declaration is reported beside the level rather than hidden inside it.