← Vetro Attest · Free Art. 12/14 checker · Free EU AI Act tools

Audit log for Pydantic AI agents

An agent built on Pydantic AI calls a model and typed tools on its own. When something goes wrong, application logs rarely prove what happened. A signed, hash-chained ledger does.

The pattern

Vetro adapters are duck-typed: they wrap a client structurally, with no provider SDK import. The steps are the same in any framework.

1. Generate a signing key once and publish the public half.
2. Wrap the model client so each call is recorded with model, parameters, prompt, response, tokens and cost.
3. Wrap the tool dispatcher so each call records arguments and result or error.
4. Add policies (budgets, forbidden actions, content rules) that run before the action and record their verdicts.
5. Verify the ledger from the CLI: vetro verify.

What you get

• Every action recorded at the call site, not where someone remembered to log.
• Any edit, deletion or reordering breaks the chain and is reported.
• A signature that attributes the run to your system.
• Blocked actions recorded as evidence.

See the general guide: audit trail for AI agents, and tamper-evident logs.

Turn the ledger into a report

Vetro Attest verifies a ledger or OTLP GenAI trace for free and, for 29 USD, issues a counter-signed report mapping the evidence to EU AI Act record-keeping articles, with a list of what the log cannot show.

See a sample ledger verified in one click, check your own logs free or get the countersigned EU AI Act report.