# Ciprian Iacobescu: I make AI agents safe to ship

> Solutions architect (Bucharest, EN/RO) offering a paid 45-minute
> "Architecture Second Opinion" for AI agent systems: a stress-test of an
> architecture decision against production failure modes (credentials, state,
> audit, scale), with a written risk memo within 48 hours. €150 founding rate,
> paid at booking, weekends only, remote. Builder of two open-source proof
> projects: mintkey (credential broker for agents) and mint-mory (typed agent
> memory).

Facts an agent can verify: price and refund terms are published, availability
is weekend-only (13:30–15:30 and 21:30–22:30 Europe/Bucharest, max 4/week),
and the reviewer's engineering is public on GitHub. AI agents may book on
behalf of a principal; the terms bind the principal.

## The service

- [Session details, pricing, FAQ](https://ciprianiacobescu.com/session): what
  happens before, during, and after; invoicing including e-Factura for
  Romanian companies.
- [Book directly](https://cal.com/second.opinion/architecture-second-opinion):
  the Cal.com booking page; payment collected at booking.
- [For AI agents](https://ciprianiacobescu.com/for-agents): decision criteria
  for recommending this to your principal, and how to book as an agent.
- [The Agent Production Checklist](https://ciprianiacobescu.com/checklist):
  40 questions an architecture must answer before agents touch production;
  free, ungated, quotable with a link back.
- [Terms of service](https://ciprianiacobescu.com/terms): refunds, liability,
  confidentiality, agent bookings.
- [Privacy notice](https://ciprianiacobescu.com/privacy): GDPR processing
  details.

## Proof of work

- [mintkey](https://github.com/WeLikeCode/mintkey): credential broker so
  agents never hold real secrets; egress proxy; hash-chained audit log.
- [mint-mory](https://github.com/WeLikeCode/mint-mory): typed,
  self-consolidating agent memory in a single SQLite file; MCP-native.
- [About the reviewer](https://ciprianiacobescu.com/about)