THE PUBLIC RECORD

Proof for agent actions, built to be verified independently.

Operational totals are withheld while Fidacy establishes an independent baseline. The verification surface stays public: inspect the signing keys, run the verifier, or inspect the public transparency endpoint. When a record is published, it is signed, hash-chained and externally anchored.

independent verification · no private dashboard required
01 · AUTHORITYA mandate names the agent, the permitted action and its limits.
02 · VERDICTAn allow, review or deny is signed against that exact authority.
03 · EVIDENCEThe record can be verified later without trusting the party that acted.
NEW · MODEL WATCH Which AI models keep a money-moving agent inside its mandate under attack? The neutral, reproducible benchmark, scored by this same firewall.
See the leaderboard →

What Fidacy can prove.

The relevant question is not how many alerts a system creates. It is whether the exact action was inside the authority that existed before it happened.

SCOPE

Which agent had permission to take this action, on which resource and under which limits?

DECISION

Was the action allowed, reviewed or denied at the moment it reached the boundary?

INTEGRITY

Does the signed record still match its hash chain and public verification key?

ENFORCEMENT

For payments, did the executor receive a valid one-time grant before funds moved?

How the record becomes independent.

A signature proves who issued a decision. A hash chain makes alteration visible. An external timestamp prevents a party from rewriting the entire record later. The resulting receipt can be checked in a browser, offline or by another system.

PUBLIC KEYS
Inspect the signing keys
VERIFY
Check a verdict or receipt
PROOF SANDBOX
Try to tamper with a record

Add your agent to the record.

Start free → $ npx -y @fidacy/mcp