PROVE WHAT YOUR AI AGENT DECIDED, AND WHY.

Every time your agent answers or escalates, Query Fanout saves a tamper-evident record of the decision and the sources behind it.

> QUICKSTART> SEE WHAT AN AUDITOR SEES
>> ONE REAL DECISION RECORD
!! OUTPUT FROM RUNNING THE CODE / support policy
QUESTION
How often does the Bitcoin block reward halve?
DECISION
ANSWER · confidence 74% · “All policy checks passed.”
SOURCES
  • developer.bitcoin.org · authority 90% · counts
  • en.wikipedia.org · authority 84% · counts
  • research-agent-7 (another AI agent) · its citation was checked against the Wikipedia page sent with it, so it counts once, not twice
PROOF
record hash 8251b16d… · Ed25519 signature valid

A question that needs live data, like “What is the Bitcoin price right now?”, escalates instead. Make this call yourself →

WHAT EACH RECORD CAPTURES

Why the agent answered or escalated: the question, every source and its score (sources from other AI agents too), and the reason.

Every reason →

PROVE EVERY DECISION

The database refuses edits and deletes. Each record is numbered, signed and chained to the one before, so a changed or missing record fails verification: a gap is itself evidence. Check it yourself, offline, with qf-verify.

How the ledger works →

HAND AUDITORS THE PROOF

Export a signed evidence pack for any decisions. Reviewers check every record, signature and checkpoint themselves, without trusting us.

Download one from the demo →

More: How it works · Compliance mapping · API reference